An independent, diaspora-endowed body to train and certify the adults who run Bangladesh's residential schools, madrasas and orphanages — and to certify the institutions themselves against a published residential child-safety standard.
Hundreds of thousands of Bangladeshi children sleep tonight in institutions that no authority inspects, run by adults no one has vetted, under a discipline regime the law still permits. This draft does not ask the state to fix that. It asks Bangladeshis abroad to build the thing the state has never built, and to make it worth having.
The Amanat Council is a non-governmental certification body. It would be endowed and initially governed by the Bangladeshi diaspora, staffed and progressively governed inside Bangladesh, and it would issue two things and nothing else: a certificate to a person who has passed a rigorous training and vetting pathway, and a certificate to an institution that meets a published residential child-safety standard. It licenses nobody. It closes nobody. It publishes two registers and lets families, donors and imams read them.
Amanat means a trust placed in your hands, to be returned intact. A child handed to a residential institution is the clearest case of it there is. The Council's standards are drafted as an expression of that obligation, co-authored with Bangladeshi ulama, not as an audit imported from Geneva. The evidence from Indonesia is unambiguous on this point: child protection in religious boarding schools fails when it arrives as external state compliance and holds when it is grounded in the tradition's own normative language.40
Read the first two figures together. Violence against children is close to universal in Bangladeshi childrearing — it is the ambient norm, not fringe behaviour by bad institutions — but only about a third of adults actually think it is necessary. The practice has outlived the belief. That gap is not a moral failing to be denounced; it is an operational opening. Most adults hitting children in Bangladesh are not committed to hitting children. They have never been shown another way to run a room of forty boys at nine o'clock at night.
It is not a proposal to reform the madrasa curriculum, to secularise religious education, or to bring Qawmi institutions under state control. Every one of those has been attempted in South Asia and every one of them converted a child-protection question into a sectarian one, with the children losing.4647 This document is about whether a nine-year-old is safe in a room at night. It takes no position on what he studies in the morning.
Bangladesh has won the enrolment argument and lost the learning one. It has almost closed the gender gap in schooling and barely touched the violence gap. And on the specific question this document exists to answer — how many children live in institutions, and are they safe — the state does not know.
This matters because it defines what the remaining problem is. Bangladesh gets children into school. It does not keep them learning, and it does not keep them safe.
| Indicator | Value | Year | Source |
|---|---|---|---|
| Total population | 175,686,899 | 2025 | World Bank4 |
| Population aged 0–14 | 48,569,949 | 2025 | World Bank4 |
| Primary completion rate | 94.3% | 2024 | World Bank4 |
| Lower secondary completion rate | 73.8% | 2024 | World Bank4 |
| Upper secondary completion rate | 31.3% | 2022 | UNESCO UIS4 |
| Children out of school, primary | 918,575 | 2024 | World Bank4 |
| Adolescents out of school, lower secondary age | 23.9% | 2024 | World Bank4 |
| Out-of-school rate, upper secondary age | 28.5% | 2024 | UNESCO UIS4 |
| Adolescent boys out of school (lower secondary) | 15% | 2024 | UNESCO GEM5 |
| Adolescent girls out of school (lower secondary) | 4% | 2024 | UNESCO GEM5 |
| Primary net enrolment, administrative census | 97.6% | 2022 | Annual Primary School Census7 |
| Primary net attendance, household survey | 84.3% | 2025 | MICS 20252 |
| Lower secondary net attendance, household survey | 59.6% | 2025 | MICS 20252 |
| Upper secondary net attendance, household survey | 50.5% | 2025 | MICS 20252 |
The two measures disagree, and the disagreement is the story. The administrative census, compiled from what institutions report, puts primary net enrolment at 97.6 per cent. The household survey, which asks families, puts net attendance at 84.3 per cent. Thirteen percentage points of Bangladeshi childhood sit between what institutions say and what families say. This document is about the part of the education system that reports on itself and is never independently checked, and the gap between those two rows is the first evidence that such a system exists.
The gender pattern has also inverted. Bangladesh is now one of the few countries in the world where adolescent boys are markedly more likely to be out of school than girls. Those boys are not nowhere. A large share of them are in religious residential institutions, which is precisely the population this policy concerns.
Boys do worse than girls here too: 53 per cent of boys fail to reach minimum reading proficiency at the end of primary school, against 48 per cent of girls.3 The national assessment reaches the same conclusion from Bangladeshi data — roughly half of primary pupils cannot read Bangla at grade level and around two thirds cannot perform basic arithmetic.6 The 2025 household survey puts foundational reading among children aged 7 to 14 at 50.1 per cent and foundational numeracy at 39.2 per cent; numeracy has improved sharply since 2019, reading has barely moved.2
None of these instruments samples by residence. The national assessment samples recognised primary schools. The household survey samples households, and by construction under-covers children who live where they study. There is no measurement of learning, and none of safety, for children in Qawmi hostels, lillah boardings, orphanages or cadet colleges. They are missing from the denominator of every number on this page.
Education spending has sat around two per cent of GDP for two decades, against the four to six per cent benchmark in the Education 2030 Framework, and the FY2025–26 allocation was estimated at about 1.5 per cent of GDP.455 Roughly a fifth of primary teachers and nearly half of secondary teachers are not recorded as trained.4 No teacher in the Qawmi system is counted in that statistic at all, because the Qawmi system is not counted.
| Finding | Value | Year | Basis | Source |
|---|---|---|---|---|
| Children 1–14 experiencing physical punishment or psychological aggression | 85.7% | 2025 | National household survey (prevalence) | MICS 20252 |
| Same indicator, previous round | 88.8% | 2019 | National household survey (prevalence) | MICS 20191 |
| Children aged 5–9 experiencing it — the peak age band | 89.8% | 2025 | National household survey | MICS 20252 |
| Adults who believe physical punishment is necessary | 32% | 2025 | National household survey (attitude) | MICS 20252 |
| Children reporting corporal punishment at school | 83% | 2013 | Ministry baseline study, n=1,210 | MoWCA / BLAST15 |
| Children reporting corporal punishment in residential institutions | 68% | 2013 | Ministry baseline study, n=1,210 | MoWCA / BLAST15 |
How to read these numbers. The MICS figures are nationally representative household-survey prevalence and describe caregivers, not institutions. The 2013 ministry baseline is the only Bangladeshi study located that separates settings; its sample is small and it is now thirteen years old, so it is cited as indicative, not as a national rate. Figures published by human-rights monitors such as Ain o Salish Kendra are tallies of reported incidents compiled from newspapers and are not prevalence at all.18 This draft does not mix the three, and no reader of it should.
Corporal punishment of children in Bangladesh remains lawful in alternative care settings, in day care, and in penal institutions. The 2011 High Court judgment and the associated ministry circular reached schools; section 89 of the Penal Code 1860 still shelters the rest, and the Children Act 2013 contains no general prohibition.1415 A residential madrasa hostel sits in the gap. Whatever else is disputed, this is not: there is no Bangladeshi law that clearly forbids beating a child in a lillah boarding at midnight.
This is the single most important fact in the document. Bangladesh cannot produce a defensible national count of the children living in residential educational care. The estimates that exist differ by a factor of four.
| Setting | Institutions | Children | Basis and year |
|---|---|---|---|
| Qawmi madrasas — only official survey ever conducted | 13,902 | 1,398,252 | BANBEIS national survey, 2015; 73,731 teachers7 |
| Qawmi madrasas under one board (Befaq), as claimed by that board | ~29,000 | ~5,000,000 | Befaq official quoted in press, 2025 — a claim, not a count8 |
| Qawmi students — range in the academic literature | — | 1.4m – 5.2m | Asadullah & Chaudhury 2016; Barkat et al. 201111 |
| Candidates sitting the Befaq board examinations | — | 349,776 | 2025, up 55% from 225,631 in 20228 |
| Registered private orphanages (etimkhana) receiving the capitation grant | 4,209 | 116,666 | Department of Social Services, FY2025–269 |
| Actual residents of those same orphanages — derived, because the grant covers at most 50% of boys and 75% of girls | 4,209 | 155,000 – 233,000 | Inference from the Department's own cap rule9 |
| Government children's homes (Shishu Paribar) | ~85 | ~9,000 | Department of Social Services11 |
| Cadet colleges | 12 | no published total | Bangladesh Army; fully residential4 |
| The only national estimate of children in residential care ever produced | — | >49,000 | UNICEF, 2008; explicitly excludes madrasa residents11 |
| Orphaned or highly vulnerable children in Qawmi madrasas — estimate only | — | 210,000 – 780,000 | 15% orphan share (Barkat et al. 2011) applied to the 1.4m–5.2m range11 |
The national education statistics agency runs an annual census of every institution in the country. It records whether each has electricity, computers, drinking water, separate toilets, solar power and internet. It does not ask whether the institution has a hostel, or how many children sleep in it. Not for schools, not for madrasas, not for colleges.7
The only official Qawmi figures in the current statistical yearbook come from the 2015 survey and have been reprinted, unchanged, in the 2022, 2023 and 2024 editions, still carrying the 2015 footnote.7 The state has had no updated count for over a decade — across exactly the period in which Dawra-e-Hadith gained degree equivalence and Al-Haiatul Ulya was created.
The cheapest recommendation in this document. Add two questions to the existing annual census: does this institution provide residential accommodation, and how many students reside on the premises. Two fields on a form the agency already sends to every institution in Bangladesh would create a national residential census at close to zero marginal cost. It requires no new law, no new budget line and no confrontation with any sector. It should be done whatever happens to the rest of this proposal.
The derived rows are inferences, presented as such. It is nonetheless the most consequential number in this document: the number of children in Bangladesh who are simultaneously parentless or destitute and living full-time in an institution that no authority inspects. The state's own residential care capacity, at roughly nine thousand places, is between one and nine per cent of it.
Bangladesh's national encyclopaedia defines two kinds of orphanage: the independent orphanage, and lillah boarding — "an orphanage attached to a madrasah".10 This is the institution at the centre of this policy, and it deserves to be described accurately rather than denounced.
A lillah boarding takes children whose families cannot feed them — overwhelmingly the children of landless and day-labouring households, and of parents who have both migrated for work — and gives them food, a bed, clothing and an education, free, funded by zakat and neighbourhood donation. It asks nothing of the family. It turns nobody away for being poor. It is, in the plainest sense, Bangladesh's largest child welfare system, and it was built entirely by poor people's charity while the state's residential capacity stalled at nine thousand places.
It is also, structurally, the most dangerous configuration in child protection: a closed institution, isolated from external scrutiny, holding the most powerless children in the country, run by a single figure of unchallengeable religious authority, with adults sleeping in the same building as children and no complaint route that does not run through the accused. Every element of that sentence is drawn from the international evidence on where institutional abuse concentrates.3334 Both things are true at once. A policy that refuses to say the first will never be listened to about the second.
Qawmi madrasas take no state money and want none. They are financed by zakat, sadaqah, qurbani hides, donation boxes and remittances from the Gulf and the United Kingdom. That independence is the source of both their reach and their opacity: an institution that receives nothing from the state owes the state no return, and no Bangladeshi body has ever had a lawful basis on which to enter one and look at the sleeping arrangements.11
This cuts the other way too, and it is the hinge of the entire proposal. An institution financed by voluntary giving is exquisitely sensitive to the confidence of the people who give. It cannot be regulated. It can be certified, if certification is something its donors come to expect.
Bangladesh does not lack child protection law. It lacks any body with the authority, the mandate and the capacity to walk into a residential institution and look.
| Setting | Registering body | Power to inspect the hostel | Binding child-safety standard |
|---|---|---|---|
| Qawmi madrasa with hostel / lillah boarding | None. The six Qawmi boards and Al-Haiatul Ulya accredit examinations, not premises | None | None |
| Alia madrasa with hostel | Bangladesh Madrasah Education Board | Academic and examination oversight; no residential welfare mandate | None specific to residence |
| Registered private orphanage (etimkhana) | Department of Social Services, under the 1961 Voluntary Social Welfare Agencies Ordinance, whose inspection power is limited to financial matters13 | Yes, tied to the capitation grant; capacity is the binding constraint | Grant conditions; no published residential care standard |
| Government children's home (Shishu Paribar) | Department of Social Services | Yes, direct line management | Departmental rules |
| Cadet college | Ministry of Defence | Internal, through the governing body | Internal regulations, not public |
| Private boarding school / English medium with hostel | Registration as a school only | None over the residential function | None |
| Coaching-centre hostel, mess, informal boarding | None | None | None |
Compiled from the Children Act 2013 and its authoritative commentary, the Voluntary Social Welfare Agencies Ordinance, the Qawmi Madrasah Sanad Act 2018, and Department of Social Services practice.149 The 2018 Act granted the Dawra-e-Hadith degree equivalence to a master's degree. It regulates a certificate. It says nothing about a dormitory.
The Orphanages and Widows' Homes Act 1944 — a statute older than the country, still on the books — carries an unrestricted power of entry into a residential child institution, and that power is expressly delegable to a person who is not a government officer.12 Its maximum penalty is five hundred taka, which is why nobody uses it. Section 64 of the Children Act 2013 likewise makes inspection delegable.14
This matters more than the penalty suggests. It means a lawful route already exists by which a non-government body could be authorised to enter and inspect a residential child institution, without new primary legislation and without creating a state inspectorate. The Council does not need this to operate, because its scheme is consensual. But it is the mechanism by which a future government could give the Council's audits legal force cheaply, and it should be preserved in any redrafting of the statute book.
The state should, in time, legislate a residential standard. This draft assumes it eventually will, and is designed to hand it a working one when it does. But a statutory scheme requires an inspectorate, and an inspectorate requires money the state does not have, in a sector that receives no state money and would resist state entry as a matter of principle. India shows what mandatory registration without inspection capacity actually produces, and it is examined in §3.
Six jurisdictions were examined against a single question: what actually changed the behaviour of a religious or residential institution that the state could not or would not close?
The Royal Commission into Institutional Responses to Child Sexual Abuse (2013–17) remains the largest evidentiary base in the world on how institutions fail children.30 It produced the ten National Principles for Child Safe Organisations, written deliberately so that a small, poorly resourced, faith-based organisation could implement them.31 Victoria converted them into eleven mandatory Child Safe Standards.32 Australia also publishes a voluntary national boarding standard, AS 5725, which is the closest existing model for a residential standard that is published rather than legislated.35 The Commission's own commissioned research is the analytical backbone of this draft: Palmer showed that organisational culture, not individual pathology, is the variable that determines whether abuse is detected and stopped.33
Taken: the ten-principle architecture and the insistence that children themselves must be able to speak. Left: the enforcement machinery, which presumes a functioning regulator, mandatory reportable-conduct schemes and universal working-with-children checks.36
England's National Minimum Standards for boarding schools is the most directly usable document found in this review, because it legislates precisely the things that go wrong in a Bangladeshi hostel.20 Standard 20.7 requires at least one staff member sleeping or on duty in every building. Standard 20.9 requires staff accommodation to be separated from boarders'. Standard 20.10 forbids boarders from entering staff residential accommodation and requires one-to-one situations to be avoided. Standard 11.4 requires an identified adult outside the staff and the leadership whom a child may contact directly. Standards 14.1–14.3 require a complaints route for boarders themselves, with a written record and a guarantee that no child is penalised for complaining. Standard 19.2 requires criminal-record and barred-list checks on every person aged sixteen or over living on the same premises as boarders. Above them sits the multi-agency framework of Working Together to Safeguard Children.22
The public inquiry into child sexual abuse concluded that boarding schools "could be said to provide the ideal environment for grooming", and recommended that the state "set nationally accredited standards and levels of safeguarding training" and make the highest level mandatory for heads and proprietors.26 Its final report remains the fullest account anywhere of how institutions protect themselves rather than children.24 That recommendation is, almost exactly, what this document proposes to build for Bangladesh. The same inquiry found that among mosques reviewed by one specialist consultancy roughly 60 per cent had no child-protection policy at all, and that of 311 faith institutions claiming to have a policy, more than 150 could not produce a current, signed-off one.25
England tried exactly what is proposed here, for madrasas, and it did not work. After consulting on registering out-of-school settings in 2015–16 and receiving over 15,000 representations, the government abandoned the model in 2018 and funded voluntary local accreditation pilots instead. The published evaluation found that nearly all participating authorities had intended to build a safeguarding accreditation scheme or kite-mark and nearly all abandoned it, citing the absence of consistent measures of safety, the cost of annual re-verification, and their own liability once they had declared a setting safe. Where marks survived, uptake was four or five settings per authority, and the authorities "were unable to evidence that accreditation affected the choices or behaviours of parents". Only already-motivated providers took part.23 The voluntary code of practice published in 2020 remains non-statutory in 2026, and research on British madrasas had identified the same governance and recruitment weaknesses a decade earlier.2928
The design consequence is the most important in this document. A voluntary certificate that relies on institutions wanting it will certify the institutions that were already safe. The Amanat Council is therefore built demand-first: the certificate must be something a donor requires before releasing zakat, a parent asks for before handing over a child, and an imam mentions before recommending a madrasa. Every element of §10 exists to manufacture that demand. Supply-side voluntarism is a proven failure and is not the plan.
Indonesia has roughly the same institution Bangladesh has, at greater scale, and has spent six years regulating it with the sector's consent rather than against it. Law No. 18 of 2019 on Pesantren gave Islamic boarding schools statutory recognition, degree equivalence and access to public funding in exchange for registration and minimum standards.37 The Ministry of Religious Affairs then issued a binding regulation on the prevention and handling of sexual violence in religious education units in 2022,38 followed in 2024 by a directorate-general regulation on child-friendly care inside pesantren, and a national "child-friendly pesantren" programme with published indicators.39
The most recent systematic review of this literature — fifty-one peer-reviewed studies, PRISMA protocol, published March 2026 — reaches a conclusion this draft treats as foundational: child protection in pesantren is compromised less by weak enforcement than by a normative gap between Islamic legal principles and state regulatory instruments. The authors score the coherence deficit at 9.75 out of 12 and argue that effective child protection "requires principled coherence between Islamic law and state regulation, not compliance with the latter alone".40
Indonesia also supplies the delivery mechanism. Kuliah Kerja Nyata places undergraduates in structured community service for academic credit as a condition of graduation, at national scale, and has done so for decades. It is the working proof that the university-student fellowship proposed in §7 is administratively possible in a South and Southeast Asian public university system.
Regulation of religious residential institutions in both countries followed catastrophe rather than preceding it. The 2017 fire at the Darul Quran Ittifaqiyah tahfiz centre in Kuala Lumpur killed twenty-three people, most of them children, in a building that was not registered with the education ministry.41 Turkey's Diyanet issued a detailed directive governing Quran courses and their student dormitories after comparable scandals.42
The lesson is about sequencing and it is grim: in a faith-education sector, the political window for safety regulation opens after children die and closes within about eighteen months. Bangladesh has had its openings — the Feni case of 2019 among them — and let them close. A standing body that already exists when the next one opens can convert it into permanent change. A committee formed in response cannot.
India requires every institution housing children in need of care and protection to register under section 41 of the Juvenile Justice Act 2015, on pain of criminal penalty, with quarterly inspections mandated.43 The national child rights commission then audited 7,163 child care institutions. The results are the strongest available argument against copying this model into Bangladesh.
The last figure is the diagnostic one. Where inspection is mandated but under-resourced, it does not fail loudly — it produces a near-perfect pass rate and a paper trail of compliance. School hostels fall outside the Act altogether, so the commission issued non-binding guidelines for them instead.45
India also supplies the cautionary tale about politics. When the same commission recommended withdrawing funding from madrasas, the Supreme Court upheld the state madrasa law and the initiative collapsed.46 Child protection framed as an attack on a religious community loses, and takes the children's interests down with it.
After the 2014 Peshawar school attack, Pakistan negotiated for three years with its five madrasa federations and produced a 2019 settlement under which registration moved to the education ministry. A new Directorate General of Religious Education had registered 17,738 madrasas by September 2024. The settlement then collapsed: a 2024 amendment reverted registration to district commissioners, granted madrasas immunity from registration under any other law, and required registration, annual activity reports and audited accounts — and contained no child-safeguarding requirement of any kind.47
Six years of the most intensive madrasa reform effort in the region produced a statute about ideology and money that is silent on the children boarding inside. When madrasa regulation is driven by security or finance, child safety is not merely deprioritised. It never enters the text.
A certifier of people and institutions. Not a regulator, not a funder of madrasas, not a curriculum reformer, not an NGO delivering services. Its entire product is a credible signal, and its entire asset is the credibility of that signal.
Two entities, deliberately separated so that neither a Bangladeshi political shift nor a diaspora factional split can capture the whole.
| Entity | Where | Holds | Cannot |
|---|---|---|---|
| Amanat Trust | Registered abroad, in a common-law jurisdiction with a strong charity regulator — a UK charitable incorporated organisation, with a US 501(c)(3) friends-of entity for American donors | The endowment; the trade marks and the seal; the standards as intellectual property; the assessor register; the appeals panel of last resort | Operate in Bangladesh; employ Bangladeshi staff; hold Bangladeshi assets |
| Amanat Council Bangladesh | Bangladesh, as a company limited by guarantee under s.28 of the Companies Act 1994, registered with the NGO Affairs Bureau for foreign-donation clearance | All operations: training, assessment, the two registers, the helpline, staff | Own the seal; change the standards unilaterally; receive unrestricted foreign funds outside the approved route |
Foreign funding runs through the Foreign Donations (Voluntary Activities) Regulation Act 2016 and NGO Affairs Bureau project approval.19 This is a real constraint with real delay, and §10 budgets for it rather than assuming it away. The split structure exists precisely so that a suspension of the Bangladesh entity's registration does not destroy the standards, the registers or the endowment.
A published list of prohibitions is a governance instrument. It is what allows a suspicious muhtamim to be told precisely what cannot happen to him.
The Council has no legal power. Its authority rests on four things it must earn in order: the public endorsement of respected ulama; a training product good enough that teachers recommend it to each other; a register that families and imams actually consult; and donors — zakat committees, diaspora mosques, Gulf and Turkish foundations, international Muslim charities — who begin to ask for the certificate before they give. If the fourth does not happen, the scheme becomes what England's did: a mark held by institutions that were already safe. §10 treats building that demand as the primary programme, not as communications.
The strongest argument for this proposal is that Bangladesh's own religious scholars published most of it first, six weeks before this draft was written, and asked for exactly the instrument the Council would provide.
On 28 July 2026, Sheikh Ahmadullah — writing as "a former Qawmi madrasa student, teacher and well-wisher" — issued a six-point public appeal on preventing child abuse in madrasas. He opens by defending the sector as one of the country's social safety nets, carrying the food, clothing, housing and medical costs of lakhs of orphaned and destitute children through lillah boarding and etimkhana without a taka of state money, and then states plainly that covering up physical and sexual abuse is in no way acceptable.49 His six points are, in substance:
| Sheikh Ahmadullah's proposal, July 2026 | Where it appears in this draft |
|---|---|
| A central investigation committee under Al-Haiatul Ulya combining senior ulama with legal and psychological experts, empowered to refer cases to law enforcement, to maintain a national blacklist so an abuser "never again gets a job in any madrasa in the country", and to publicly clear the falsely accused | The Barred Register and the Clearance Ruling, §5 |
| A complaints cell with a hotline, strict confidentiality, and psychological support to the victim's family | Standard S8 and the Council helpline, §6 |
| CCTV in classrooms, verandas and especially the Nurani, Hifz and residential sections; a mandatory minimum one-cubit gap between beds; separate cots where funds allow; an absolute prohibition on teachers taking personal service from students | Standard S5, verbatim, including the one-cubit rule |
| A strict written code of conduct; regular teacher training in child psychology, Islamically grounded gentleness with young learners, and child-protection law; staff supervising young children to be married, of mature age, with family quarters and regular leave | Standard S4 and the Level 2 curriculum, §5 |
| Girls' madrasas staffed and administered entirely by women, all guardian contact through female staff | Standard S4.6 |
| Discouraging residence for very young children, because institutional living deprived of parental affection leaves them psychologically weak and unprotected | Standard S1.4, the presumption against residence under ten |
| "If the boards impose these conditions for registration and monitor strictly, the problems will largely diminish" | The entire mutual-recognition strategy, §9 milestone M2 |
He is not alone. In 2019, after five madrasa teachers were arrested on rape charges in a single month, a Hefazat-e-Islam spokesman said the organisation had told a conference of 1,200 principals to take a tough stand, and a pro-madrasa editor and former madrasa teacher publicly called for closed-circuit cameras in students' sleeping quarters.50 In 2026 Mufti Habibullah Mesbah put the case for the register in one sentence: "Simply expelling a teacher from one institution achieves nothing. They just migrate to the next village. We need a centralised, binding registry run by the Ulema. If a teacher is accused of a moral crime, their licence must be permanently revoked nationwide."51
The Amanat Council is not a foreign intervention into Bangladeshi religious education. It is the administrative machinery for a reform that senior Bangladeshi ulama have already publicly demanded and that the boards have said they lack the capacity to build. The diaspora's contribution is money, technical standards and the one thing a domestic body cannot easily supply — an assessor who owes nothing to the institution he is assessing.
The framing follows from this. The Council does not arrive saying your institutions are dangerous. It arrives saying you asked for a blacklist, a hotline, a code of conduct and trained teachers; here is the body that can run them, and you hold a veto over how it speaks.
Leadership is split. In the same period Befaq has dismissed abuse allegations as "anti-Islamic propaganda", and a Hefazat nayeb-e-ameer who is also Befaq's secretary general could not say how many teachers the board had ever disciplined, stating that cancellations happen "from time to time" and exact numbers "cannot be stated".52 Investigative reporting in August 2026 documented a teacher accused of sexually abusing students who simply opened another madrasa, because no registry exists that would have stopped him.51
The state has tested the alternative route three times — in 2006, 2013 and 2016 — and retreated each time. In October 2013 a draft law creating a Qawmi Madrasa Education Authority with government representatives was due before cabinet; the day before, the Hefazat chief warned of civil war, and the ministry withdrew it.53 Any design that requires the state to compel this sector is a design that has already failed three times.
Three findings from the research literature shape every design decision that follows, and all three are counter-intuitive to the people who usually fund this kind of work.
The Australian Royal Commission's literature review, drawing on more than four hundred documents, concluded that "the best way to reduce the risk of institutional child sexual abuse is to avoid dangerous practice rather than attempt to screen out allegedly dangerous people".34 In the core sample of one study of people who abused inside organisations, only one of twenty-one offenders had a relevant previous conviction.
The implication for this policy is uncomfortable and important. A barred register is necessary and nowhere near sufficient. It catches the person who has already been caught. It does nothing about the adult who has never been reported, which is most of them. This is why Standard S5 — who sleeps where, how far apart the beds are, whether an adult can be alone with a child behind a closed door, who is awake at night — carries more weight in this scheme than Gate 1 does. Palmer's companion report explains why residential religious institutions sit at the top of the risk hierarchy: they approximate the "total institution", with an alternative moral universe, extinguished prior identities, secrecy and a unique power structure, each of which independently facilitates abuse, slows its detection and impedes the response to it.33
Fifty-six per cent of adolescents in Israeli residential care reported peer physical violence and seventy-three per cent verbal victimisation within a single month.59 The Royal Commission's review notes that children in out-of-home care may be at higher risk from peers than from adult staff. England's boarding standards now require an explicit policy on child-on-child abuse naming "the risks associated with children sharing overnight accommodation".20
Almost every conversation about madrasa safety in Bangladesh is a conversation about teachers. The evidence says that a scheme which addresses only adult-to-child harm will miss the larger share of what is happening in a room where forty boys sleep, and will be judged by families to have missed it. Peer abuse is therefore named explicitly in Standard S4, in the S5 sleeping arrangements, in module A2 and in the supervision ratios, rather than being folded into a general duty of care.
This is the finding that makes the entire proposal defensible rather than hopeful.
Tanzania's Interaction Competencies with Children for Teachers programme — a five-and-a-half-day workshop — produced large reductions in teacher-reported violence and, notably, reduced peer-to-peer violence as a secondary effect. The same method has been used with caregivers in twenty-four Tanzanian orphanages, which is the setting closest to a Bangladeshi etimkhana.61 These are randomised trials in low-income countries with residential and resource conditions comparable to Bangladesh's, and they are the evidential basis for module B2 and for the length and format of the Level 2 practicum.
One caveat belongs here rather than in a footnote. The Ugandan trial found no effect on educational test scores or on children's mental-health scores.60 This policy therefore claims that a well-designed programme can reduce the violence adults do to children in institutions. It does not claim that doing so will improve learning outcomes, and any version of this proposal that promises a minister better examination results is overselling the evidence.
Children tell friends first, then family, then adults at their institution. They do not report because they fear being called a snitch, because they believe nothing will be done, and because they do not know what happens next. Four design consequences follow, and each is written into Standard S2. Publish, in Bangla and on the wall, what happens after a complaint. Legitimise peer reporting rather than punishing it. Require several trusted adults per institution rather than one designated officer, because a child who does not trust that particular adult has nowhere else to go. And install an outside visitor.
That last one carries unusual evidential weight. In the Royal Commission's review of out-of-home care, "outside visits by a trusted adult authority" was one of only two protective factors the literature could identify at all.34 It is the reason the University Amanat Fellow described in §7 is not merely a training facilitator: the fellow is formally designated as the institution's external visitor, named on the wall, returning on a schedule, known to the children by face. A recurring outsider with no stake in the institution's reputation is, on the evidence, one of the few things that reliably helps.
Two operational warnings from jurisdictions that have done this. In Victoria's reportable conduct data, 41 per cent of allegations were physical violence, not sexual abuse. A scheme designed around sexual abuse will be overwhelmed by the far larger volume of ordinary beating, and will handle it badly.36
And in Western Australia, introducing a mandatory reporting duty multiplied reports 3.7-fold while substantiations only doubled.57 The lesson is sequencing: build the reporting capacity before imposing the reporting duty. Before Standard S8's 24-hour rule binds anyone, the Council must have a working triage line, a documented referral protocol agreed with police and the Department of Social Services, and a published number for how many cases it can actually carry. A duty to report into a system that cannot respond teaches children that reporting is pointless, which is worse than not asking them.
Cultures that do not permit discussion of these matters make it difficult for children to recognise grooming behaviour and deny them the vocabulary needed to disclose. This is not sex education, and it must not be presented as such. It is a short, ulama-endorsed Bangla lexicon — body boundaries, private parts, secrets that are not to be kept, who to tell — that a nine-year-old Hifz student can actually use. Drafting it with the Shura Council is a Milestone Zero task, because without it Standard S2 asks children to report something they have no words for.
The Cochrane systematic review of child-protection training for professionals rated every outcome low to very low certainty, and no included study measured actual official reports to authorities.57 Training reliably changes knowledge. Whether it changes behaviour depends on conditions the training-transfer literature specifies precisely: an explicit rule code rather than general principles, scenarios practised rather than described, goal-setting, a trained supervisor in the workplace, and real rewards and sanctions where the person actually works.
Those five conditions are the reason the Amanat Pathway looks as it does. The code of conduct in module A3 is a rule list, not a values statement. Gates 3 and 5 exist because practice and supervised probation are what the transfer evidence requires. Level 3 exists so that every institution contains a trained supervisor. And the institutional standards in §6 exist so that there is a consequence in the workplace. A pure training programme without those four supports would be, on the evidence, a way of spending money on knowledge that does not travel.
A certificate that is easy to get is worth nothing. Six sequential gates, no gate skippable, no certificate issued until the last one closes. The whole design assumption is that the difficulty of the process is the product.
| Level | Who must hold it | Online | Offline | Gates | Renewal |
|---|---|---|---|---|---|
| L1 Amanat Foundation | Every adult who works in or regularly enters a residential institution — teachers, cooks, guards, drivers, cleaners, night watchmen, volunteers, the muhtamim's household | 12 hrs | None | 1, 2, 6 | 3 yrs |
| L2 Certified Residential Caregiver | Anyone with night-time or pastoral responsibility: hostel superintendents, nazims, house parents, matrons, night supervisors, Hifz and Nurani section staff | 60 hrs | 3-day practicum | All six | 3 yrs |
| L3 Certified Institution Head & Safeguarding Lead | Muhtamims, principals, orphanage directors, cadet college house masters. Every certified institution must have at least one L3 holder who is not the head | +80 hrs | 5-day residential | All six, plus panel | 3 yrs |
| L4 Accredited Assessor / Master Trainer | Those who audit institutions and train fellows. Includes diaspora professionals conducting document review and video interviews remotely | +40 hrs | 3 shadowed audits | All six, plus calibration | 2 yrs |
The requirement that an institution's safeguarding lead not be its head is taken directly from English statutory guidance, which states that it is not appropriate for the proprietor to be the designated safeguarding lead.21 In a Qawmi madrasa, where the muhtamim's authority is effectively unchallengeable, it is the single most important structural change in this document. A complaint about the head must have somewhere to go that is not the head.
The largest religious-institution safeguarding system in the world is the United States Catholic Church's, which in a single recent audit year trained 2,328,545 adults and ran 2,320,143 background checks, with more than six million adults trained since 1998.48 The objection that a faith sector cannot be trained and vetted at scale is empirically false. The objection that doing so guarantees safety is also false, and §11 deals with it.
Every module carries three threads simultaneously: the fiqh and adab thread, taught by a Bangladeshi alim; the evidence thread, drawn from the international literature; and the practice thread, which is what the learner will actually do on Tuesday night. The religious thread is not decoration. Islamic Relief Worldwide's own child safeguarding policy is built on amanah, on Qur'an 8:27 — "betray not knowingly your trusts" — and on the legal maxim la darar wa la dirar, there shall be no harming and no reciprocating harm.49 A Bangladeshi standard can stand on the same ground.
| Module | Level | Hours | Core content | Assessed by |
|---|---|---|---|---|
| A1 · The child as amanah | L1 | 2 | Trust and obligation in fiqh; Qur'an 8:27; la darar wa la dirar; the muhtamim's answerability; why a child in your building is the strongest case of amanah there is | Discussion, taught by an alim |
| A2 · What harm looks like | L1 | 3 | Physical, sexual, emotional abuse, neglect and exploitation, described in Bangladeshi residential settings; grooming behaviours; peer-on-peer abuse in dormitories; what a frightened child actually does rather than what people expect | Scenario test |
| A3 · The rules I follow | L1 | 3 | The code of conduct line by line: no corporal punishment, never alone with a child behind a closed door, the two-adult rule, no personal service from students, no private messaging, photography rules, gender rules | Signed undertaking + test |
| A4 · If a child tells me | L1 | 3 | Receiving a disclosure; the four things never to say; the 24-hour internal report; the helpline; what happens next; confidentiality and its limits | Role-play |
| A5 · Law and consequence | L1 | 1 | The Children Act 2013 and the 2000 Act in plain Bangla; what a police case involves; the Barred Register | Test |
| B1 · Running a night | L2 | 10 | Supervision ratios and rotas; who sleeps where and why; bed spacing; lighting; bathroom and washing arrangements; the visitor log; roll call; what to do when a child is missing at 2am | Dormitory audit |
| B2 · Discipline without violence | L2 | 12 | Positive discipline for a room of forty; de-escalation; the specific moments teachers report losing control; classroom and hostel routines that prevent the incident; drawn from evaluated teacher-violence-reduction programmes in comparable settings | Observed role-play |
| B3 · The child who is not coping | L2 | 10 | Homesickness and institutional separation; bedwetting, self-harm, withdrawal, aggression; mental-health first aid; when to call a family, when to refer, when to sit still | Case assessment |
| B4 · Health, food and the body | L2 | 8 | Nutrition on a lillah budget; illness and injury records; scabies, TB and skin disease in dormitories; menstruation and dignity in girls' institutions; fire, drowning and electrical risk; emergency drills | Practical |
| B5 · Listening to children | L2 | 10 | Running a functioning child committee; the complaint box that is actually opened; weekly family contact; the independent adult; why children do not report and what changes that | Observed session |
| B6 · Records, boundaries and the practicum | L2 | 10 | Admissions and staff registers; the incident log; professional boundaries in a live-in institution; the practicum and its assessment | Portfolio |
| C1 · Leading a safe institution | L3 | 20 | Writing and owning a safeguarding policy; the managing committee; the safeguarding lead who is not the head; safer recruitment end to end; handling an allegation against your own teacher without dismissing him quietly into the next district | Written policy for own institution |
| C2 · When it has happened | L3 | 20 | Immediate response; preserving the child's position; referral to police and the helpline; supporting the family; managing the community and the falsely accused; what covering up costs everyone | Panel interview |
| C3 · Money, governance and trust | L3 | 20 | Keeping accounts a donor can read; separating the head's money from the institution's; publishing; the goraba and lillah funds; why financial opacity and child-protection failure travel together | Accounts exercise |
| C4 · Building the standard into the building | L3 | 20 | Reconfiguring an existing madrasa to meet S5 at near-zero cost; sequencing improvements on a real budget; the self-assessment and the audit | Improvement plan |
Total: 12 hours at L1, 60 at L2, a further 80 at L3. Delivered as ten-to-fifteen-minute audio-first lessons downloadable over a phone, because that is the device Bangladesh actually has; mobile accounts for roughly three quarters of web traffic in the country.54 Every module is also delivered in full in the offline practicum for candidates who cannot use the platform at all.
Ten standards, because a small institution with one paid administrator can hold ten in its head. Each is written so that compliance can be checked by looking, not by reading a file. Where a standard is borrowed, the source is named.
A safeguarding policy signed by the head and the managing committee, reviewed annually, displayed in Bangla where children and visitors can read it. A named safeguarding lead who is not the head. A presumption against full residence for children under ten, with a recorded reason where it is unavoidable.
A functioning child committee that meets monthly and keeps minutes. A private complaint route that does not pass through any accused adult. Contact with family at least weekly for every resident child. At least one named adult outside the staff and the management whom a child may contact directly.
Every adult in the institution holds at least Level 1. Every adult with night or pastoral duty holds Level 2. The head and the safeguarding lead hold Level 3. Police clearance and Barred Register checks on every adult, including every person aged sixteen or over who lives on the premises. References taken and recorded before appointment.
No corporal punishment or humiliating treatment, in any circumstance. No adult alone with a child behind a closed door. The two-adult rule for all overnight activity. No teacher takes personal service from a student. Girls' institutions staffed and administered by women, with guardian contact through female staff, phased against a published timetable because only about 9 per cent of Qawmi teachers are women. Staff supervising the youngest children to be of mature age. Photography and private messaging rules. Signed by every adult, annually.
Adults sleep separately from children, in accommodation children do not enter. At least one identified staff member sleeping or on duty in every building where children sleep, contactable by any child. A minimum one-cubit gap between beds in shared sleeping rooms, moving to separate cots as funds allow. Lighting on routes to bathrooms. Lockable, private bathing and toilet facilities appropriate to age and sex. A visitor log. Fire exits, extinguishers and a drill twice a year. A recorded maximum occupancy per room.
An illness and injury record for every child. A named referral route to a health facility. Nutrition adequate and recorded. Clean water. Provision for menstruation with dignity. Mental-health first aid capability among Level 2 staff. Medication stored and administered under control.
A published discipline policy consistent with S4. Adequate sleep and rest hours by age, recorded in the daily routine. No forced labour, no begging, no collection duty imposed on children, no domestic service to staff households. Study load appropriate to age.
Any concern reported to the safeguarding lead within 24 hours and logged. A published threshold at which the institution refers to police and to the child helpline. No quiet dismissal: an adult removed for a safeguarding reason is referred to the Council's Barred Register, and doing so is a condition of the institution keeping its certificate. Support to the child and family. A fair process for the accused, including public clearance where an allegation is disproved.
An admissions register naming each child, their family and their guardian's contact. A staff register with certificate numbers. An incident log. Annual accounts, published, distinguishing the institution's money from the head's. Because financial opacity and child-protection failure travel together, and donors will ask.
An annual self-assessment against all ten standards. An external audit by an Accredited Assessor on a three-year cycle. Unannounced spot checks on a random ten per cent of certified institutions each year. A corrective action plan with dates. Certification suspended, not merely queried, where a corrective action is missed twice.
Women are 65 per cent of government primary teachers, 19.6 per cent of Alia madrasa teachers and 9.3 per cent of Qawmi teachers.7 A standard that assumes a mixed workforce — female night supervision, a woman on every complaint panel, two adults of different sexes on duty — would be rejected as impossible rather than resisted as intrusive, and the institutions that most need it would be the first to be excluded.
So S4's female-staffing requirement for girls' institutions is phased against a published timetable, and the certification pathway includes a funded route for women into Level 2 and Level 3 in the Qawmi sector. Changing who works in these buildings is a ten-year project, and the standard says so rather than pretending otherwise.
Standards S1, S2, S3, S5 and S8 are adapted from England's National Minimum Standards for boarding schools, which specify night staffing in every building, separated staff accommodation, an independent adult outside the staff group, a complaints route for boarders themselves, and criminal-record checks on every person aged sixteen or over living on the premises.20 The architecture of ten plain-language standards implementable by a small faith organisation is Australia's National Principles.31 The one-cubit bed rule, the prohibition on personal service, the CCTV provision in residential wings, the female staffing of girls' institutions and the presumption against residence for the very young are Sheikh Ahmadullah's, adopted as written.49
| Tier | What it requires | What it costs the institution | Verification |
|---|---|---|---|
| Registered | Head has begun L3. All adults enrolled in L1. Self-assessment submitted. Admissions and staff registers exist. | Free, always | Desk check |
| Certified | All ten standards met. Every adult L1, every night-duty adult L2, head and safeguarding lead L3. Twelve months of incident logs. | Free below the income threshold; sliding fee above it | On-site audit by an Accredited Assessor, plus private conversations with children out of staff hearing |
| Certified Exemplar | Certified for two cycles with no upheld corrective action; hosts practicum placements; mentors two institutions towards Registered | Free; receives a mentoring grant | Full audit plus verified mentee progress |
Certification schemes fail at the frontline, not at head office. The independent auditor of the United States Catholic Church's charter states plainly that participation in the audit is not required, that in one reporting year only 24 of 61 audited dioceses permitted auditors to visit parishes and schools, and that self-reported data is not subject to a full audit until an on-site visit occurs.48 A scheme that reads paperwork will be given excellent paperwork.
Three counter-measures are therefore load-bearing rather than optional. Certification requires an on-site visit, without exception, at every cycle. Every audit includes private conversations with children, chosen by the assessor and not by the institution, out of staff hearing. Ten per cent of certified institutions are visited unannounced each year, and the sample is drawn centrally. An institution that refuses any of the three is not deferred; it is removed from the register, and the removal is published.
The Council has no campus and will never build one. Knowledge is delivered to a phone. Practice is delivered by university students on their own campuses and in nearby institutions. Verification is delivered by trained assessors who have no relationship with the institution they assess.
The typical Level 2 candidate is a madrasa teacher in his twenties earning between three and eleven thousand taka a month, with a mid-range Android phone, a data budget measured in taka, patchy electricity and limited formal literacy in written Bangla.11 The platform is designed for him and not for a laptop.
The offline half of the pathway is delivered by undergraduates, organised as a named fellowship with academic credit. This is not volunteering. It is a structured teaching placement with supervision, assessment and a stipend.
| Department | What the fellow contributes |
|---|---|
| Psychology, Educational Psychology | Facilitates the child-development and de-escalation modules; runs the disclosure role-plays |
| Social Work, Social Welfare | Facilitates case discussion, referral pathways, family contact and the child-committee module |
| Education, Teaching & Learning | Positive discipline, classroom and hostel routines, lesson observation |
| Law | The legal module; helps institutions write policies that match Bangladeshi law rather than an imported template |
| Public Health, Nursing | Health, nutrition, hygiene, dormitory disease, first aid, emergency drills |
| Islamic Studies, Arabic | Co-facilitates the fiqh thread; translates between the sector's vocabulary and the standard's |
| Architecture, Civil Engineering | The physical audit under S5: occupancy, bed spacing, lighting, fire routes, water and sanitation. This is the module institutions find most immediately useful, and it opens doors |
| Computer Science, Statistics | Registry operations, data quality, the public dashboard, offline data capture |
Structured, credited student community service at national scale is not speculative. Indonesia has run Kuliah Kerja Nyata as a compulsory graduation requirement for decades, carrying three credits and roughly 288 hours, with documented governance arrangements and known harassment risks that this design borrows its safeguards from.58 India's National Service Scheme and community-engagement mandate do comparable work. Bangladesh already has the raw components in the National Cadet Corps, the Rovers, and university internship rules. What is missing is a purpose worth organising them around.
Not fly in for workshops. The diaspora contribution is four specific things, each chosen because distance is an advantage rather than a handicap: the endowment, which buys independence from any single domestic funder; authorship of the technical standards and curriculum, alongside Bangladeshi co-authors; remote assessment — document review, video interviews with heads, and calibration of Bangladeshi assessors, which a professional in Birmingham or Toronto can do at seven in the morning before work; and the demand side, because diaspora mosques, zakat committees and hometown associations are among the largest single funders of Bangladeshi madrasas and etimkhanas, and are the constituency most able to start asking for the certificate before they give.
That last point is the whole strategy in one sentence. The diaspora's decisive asset is not its expertise. It is its chequebook, and specifically its willingness to attach a condition to it.
Mapped by what each one can actually do to the scheme, and what would make them do it. A stakeholder list that does not name the parties who can kill the proposal is a public-relations document.
| Stakeholder | Power over the scheme | Interest | What secures engagement |
|---|---|---|---|
| Al-Haiatul Ulya and the six Qawmi boards | Decisive. Can legitimise or destroy it | Protect the sector's autonomy and reputation; avoid state entry | A veto over religious framing, no veto over safety content; the Council does the administrative work they have said they cannot do; mutual recognition talks from M2 |
| Befaqul Madarisil Arabia (Befaq) | High. Currently the most sceptical major actor | Institutional standing; resistance to external criticism | Start with institutions that volunteer; let certified madrasas be the argument; never ask Befaq to endorse before members do |
| Individual muhtamims and principals | High in aggregate. They are the customers | Donations, reputation, enrolment, avoiding scandal | Free certification below the income threshold; the S5 physical audit, which is immediately useful; a certificate that donors ask about |
| Reform-minded ulama and Islamic scholars with public platforms | High and favourable | Already publicly demanding these safeguards | Adopt their proposals as written and credit them; seats on the Shura Council |
| Etimkhana operators and the private orphanage sector | Medium | Capitation grant continuity; donor confidence | Certification as evidence for the grant and for donors; free at entry tier |
| Cadet colleges and elite boarding schools | Low over the scheme, high as signal | Reputation; fee-paying parents | Invite them first and charge them full fees. Their certification funds the madrasas' and destroys the framing that this targets religious institutions |
| Stakeholder | Power | Interest | What secures engagement |
|---|---|---|---|
| Ministry of Social Welfare / Department of Social Services | High. Registers and inspects orphanages | Grant integrity; capacity it does not have | Offer certification data as free assurance on the capitation grant; ask for nothing in return at first |
| Ministry of Education and Ministry of Primary and Mass Education | Medium | Standards, teacher quality, the draft Education Act | Hand over the standards for free if the state ever legislates. The Council's success is the state adopting its standard |
| Ministry of Religious Affairs and the Islamic Foundation | Medium-high | Legitimacy with the religious sector; imam training reach | Co-delivery through imam training academies; Level 1 for mosque-based education staff |
| Ministry of Women and Children Affairs; Bangladesh Shishu Academy; child helpline 1098 | Medium | Child protection mandate, thin delivery capacity | Route S8 referrals to 1098; publish joint data; the Council becomes their reach into a sector they cannot enter |
| NGO Affairs Bureau | Decisive on foreign funding | Control of foreign donations under the 2016 Act | Full transparency, domestic revenue share rising each year, no political activity, a Bangladeshi-majority board on a published timetable |
| Police, Women and Children Repression Prevention Tribunals | Medium | Prosecutable cases | Better-evidenced referrals; training on receiving a child's disclosure |
| Public and private universities; the University Grants Commission | High for delivery | Student employability, community engagement, research output | Credit-bearing fellowship; co-authored research; named faculty coordinators |
| BRAC, Manusher Jonno Foundation, Ain o Salish Kendra, BLAST, CAMPE, Bangladesh Shishu Adhikar Forum | Medium | Long-standing child protection work; credibility | Independent evaluation, legal support, monitoring data; explicitly not implementation, to protect the Council's independence |
| Journalists and investigative outlets | High. They can end the scheme in a week | Accountability | Publish revocations before the press finds them. A certifier that hides a failure is finished |
| Parents, guardians and children themselves | Ultimately decisive | A child who comes home safe | A register searchable by phone and by SMS; a helpline; a child panel with the right to publish dissent |
| Stakeholder | Role | What is asked of them |
|---|---|---|
| Bangladeshi diaspora professional bodies — medical, engineering, legal, academic and teaching associations in the UK, US, Canada, Australia and the Gulf | Founding trustees, assessors, curriculum authors, endowment | Money, unpaid governance, and a fixed exit from control at year seven |
| Diaspora mosques, zakat committees and hometown associations | The demand engine | Adopt a giving policy: from a stated date, funds for residential institutions go to Registered or Certified ones. This single act does more than any training |
| International Muslim charities — Islamic Relief, Muslim Aid, Human Appeal, Ummah Welfare Trust and their peers | Funders of Bangladeshi madrasas and etimkhanas; safeguarding standard-setters in their own right | Recognise Amanat certification within their own downstream due diligence. Several already require safeguarding standards of partners |
| Gulf, Turkish and Malaysian foundations funding Bangladeshi religious institutions | Major funders, currently with no verification instrument | Use the register. They face the same reputational exposure and have no way to check |
| UNICEF Bangladesh, Save the Children, Terre des hommes | Technical partners; convening power | Standards review and evaluation. Deliberately not core funders, to keep the Council independent of the aid cycle |
| Bilateral donors with safeguarding due-diligence regimes | Indirect but powerful | Their partner-safeguarding requirements are the template for making certification a condition of money48 |
| Keeping Children Safe; the Boarding Schools' Association; the Australian Childhood Foundation; HQAI | Peer certifiers | Standard benchmarking, assessor calibration, subsidised audit expertise |
| Indonesian Ministry of Religious Affairs and the pesantren child-protection programme | The most useful single partner available | A working South-South partnership: they have solved a version of this problem in a comparable society and their documents are directly translatable |
| Academic partners — Bangladeshi and international | Independent evaluation | A pre-registered evaluation with the power to publish a negative result |
Five phases over ten years, each with a small number of falsifiable targets and an explicit stop condition. A milestone that cannot fail is not a milestone.
Bangladesh went to the polls on 12 February 2026. The BNP formed a government under Prime Minister Tarique Rahman, sworn in on 17 February; Jamaat-e-Islami became the principal opposition for the first time. Education and Primary and Mass Education are held together by Dr A N M Ehsanul Hoque Milon; Religious Affairs by Kazi Shah Mofazzal Hossain; Women and Children Affairs and Social Welfare together by Prof Dr A Z M Zahid Hossain.56
Three features of this moment shape the plan. The interim government created eleven reform commissions and none for education, which the Centre for Policy Dialogue called a form of national failure — so the standards vacuum is unusually open. A draft Education Act went to public consultation in February 2026 which would require every private institution to register and bring non-government educational institutions into the formal system through recognised certification; it is not enacted, and a body holding a working residential standard is well positioned if it is. And the law governing foreign-funded non-government organisations was liberalised in April 2026, with foreign-grant commitments in FY2025–26 at Tk 138.9 billion, the highest on record.56 The regulatory door for a diaspora-funded body is more open now than at any point in a decade.
Register both entities. Seat the board, the Shura Council and the child panel. Co-draft the ten standards with Bangladeshi ulama and put them out for open consultation in Bangla. Author and record the Level 1 and Level 2 curriculum. Build the platform and the two registers. Sign five university memoranda. Commission an independent baseline study of residential conditions in a stratified sample of institutions, published whatever it finds.
Stop condition: if no recognised alim will sit on the Shura Council, the project does not proceed to M1. It has no route to legitimacy and should return the money.
Pilot in three districts chosen for contrast, one of them with high Qawmi density. Certify the first cohorts. Run the helpline. Open both registers to public search by web and SMS. Certify the first institutions, including at least one cadet college or elite boarding school at full fee, so that the scheme is visibly not aimed only at madrasas. Publish the first annual report including every complaint received and every certificate refused.
The first hook. Map the institutional self-assessment onto the Department of Social Services capitation-grant cycle, whose deadlines are already fixed at 31 July at upazila level, 31 August at directorate and 30 September at ministry. Four thousand two hundred orphanages already complete an annual return, already hold a registration and already enrol their residents in school. A certification that rides that existing cycle is far more adoptable than a parallel scheme, and it reaches the lillah boardings first.
Stop condition: if fewer than 30 institutions reach full certification, or if the independent evaluation cannot detect a change in observed practice, the design is wrong. Redesign before scaling. Do not scale a scheme that has not worked in three districts.
Extend to all eight divisions and twenty universities. Begin formal mutual-recognition talks with Al-Haiatul Ulya and the boards, with the specific objective Sheikh Ahmadullah named: certification conditions attached to board registration. Convert diaspora mosque and zakat-committee giving policies. Secure recognition of the certificate inside the downstream due diligence of at least three international Muslim charities. Offer the Department of Social Services free assurance data on capitation-grant orphanages.
Stop condition: if no board has entered recognition talks and no major donor has adopted the certificate as a condition by the end of year four, the demand-side theory has failed. Convert to a pure training provider and stop claiming to certify.
The target is not a number of certificates. It is the point at which a parent in a village asks whether the madrasa is certified, and the muhtamim is embarrassed to say no. Hand the standards to government free of charge for adoption in any statutory scheme. Integrate the curriculum with national teacher-training institutions. Bring diaspora board control below fifty per cent, on the constitutional trigger.
Stop condition: if domestic revenue is below 40 per cent at year seven, the Council is a foreign-funded project rather than a Bangladeshi institution. Publish that finding and plan an exit.
Bangladeshi-majority governance completed. Statutory recognition of the Council as an accredited certifier under any new education or child protection law, if the state wants it. The diaspora role reduces to endowment stewardship and international quality assurance. A full independent evaluation against the M0 baseline, published in full, including whether children in certified institutions are measurably safer than children in comparable uncertified ones.
Only three primary studies worldwide have evaluated a safeguarding intervention in out-of-home care. A cluster-randomised or stepped-wedge evaluation in Bangladeshi residential madrasas and etimkhana would be a global first, and it is the strongest available answer to a donor asking why a non-government body should hold this mandate at all.
Stop condition: the ten-year evaluation is pre-registered and the Council commits in its constitution to publishing it even if it shows no effect, and to winding down if it shows none after a further three years.
The objection this proposal must survive is not that it is a bad idea. It is that Bangladesh has seen many good ideas consume large budgets and produce reports. That objection is correct as a prior, and the response is structural, not rhetorical.
In FY2025–26 Bangladeshi ministries spent only 67.5 per cent of the revised Annual Development Programme allocation, a five-year low, and 19.3 per cent of the entire year's development spending went out in June alone — the last month of the fiscal year.62 Money that is spent in the final weeks to avoid surrendering it is not money spent well.
The more instructive case is PEDP3, the flagship primary education programme. It met 60 of its 63 disbursement-linked indicator targets, a 95.2 per cent success rate across a US$6.73 billion programme, and was rated Satisfactory. The World Bank's own completion report nonetheless concluded that the national assessment results indicate that not all primary school children are learning.62
That is the failure mode this policy is designed against, and it is not corruption. It is a programme hitting almost every indicator it set itself while the thing the indicators were proxies for does not happen. A body that reports "N people trained" will reproduce it exactly. This is why the outcome measure in §12 is what children say, and why the M4 commitment is to wind down if ten years produce no measurable effect.
| Source | Role | Condition attached |
|---|---|---|
| Diaspora endowment | Core independence. Target: an endowment covering 30% of core cost by year ten, so no single funder can direct the standard | Capital only. Income spendable, principal not, except on a two-thirds trustee vote |
| Outcome-based grants from diaspora and international funders | The main growth funding | Paid per verified certified person and per institution still certified at 12-month re-verification. Not per workshop, not per participant-day |
| Fees from institutions above the income threshold — cadet colleges, elite boarding schools, large English-medium schools | Cross-subsidy | Full cost recovery plus a margin that funds free certification for the poorest institutions |
| Zakat and waqf, through a segregated fund | Domestic revenue, and the most legitimate money the Council can hold | Zakat spent only on eligible categories and only on direct service to children, never on core administration. Separately accounted, separately audited, published |
| Institutional and philanthropic grants for specific functions | The registry, the helpline, evaluation | No funder holds more than 25% of annual income |
| Government funding, Bangladeshi or foreign | Refused for the certification function | A certifier funded by a state can be leaned on by that state |
The zakat channel is not marginal. Academic estimation puts Bangladesh's potential annual zakat at between roughly 2.3 and 3.8 per cent of GDP depending on method, against a state Zakat Fund whose record collection was about Tk 10 crore.1762 Remittances reached US$35.56 billion in FY2025–26.62 The gap is not money. It is a trustworthy channel, and a certificate is one.
Major donors already require exactly this. The United Kingdom's development department obliges first-tier partners to cascade six safeguarding standards down the delivery chain; the Charity Commission requires trustees to conduct due diligence on grantees; Islamic Relief binds its partners contractually to its own child safeguarding standards.6249 A British Bangladeshi mosque that raises money for a madrasa in Sylhet is already, in principle, subject to trustee due diligence obligations it currently has no practical way to discharge.
The Council's single most valuable product is therefore not the training. It is a verifiable answer to a question a British, American or Gulf donor is already legally or contractually obliged to ask and currently cannot answer: is the institution I am funding safe for the children in it?
Planning figures, stated as such, at an assumed Tk 120 to the US dollar. They are built from Bangladeshi salary and travel realities, not from international consultancy rates, and every one is a published target the annual accounts will be measured against.
| Unit | Cost (BDT) | Approx. USD | Comparator |
|---|---|---|---|
| Level 1 certification, per person | 400 | 3 | Marginal cost: platform, invigilation, police-certificate verification |
| Level 2 certification, per person | 3,500 | 29 | Includes 3-day practicum and 6-month probation supervision |
| Level 3 certification, per person | 12,000 | 100 | Includes 5-day residential and panel interview |
| Level 4 assessor training, per person | 45,000 | 375 | Includes three shadowed audits and calibration |
| Institution audit, on site, two assessors | 25,000 | 208 | Keeping Children Safe charges about £15,000 for a first certification. That price is not available to a madrasa, which is why the model is student-facilitated and locally assessed48 |
| University fellow, per placement block | 6,000 | 50 | Stipend and travel; a fellow facilitates for roughly 25 candidates per block |
| Phase | Direct certification cost | Core cost | Total | Cost per person certified |
|---|---|---|---|---|
| M0 — Foundation, 9 months | — | $620,000 | $620,000 | — |
| M1 — Pilot, 15 months | $61,000 | $1,180,000 | $1,241,000 | $477 |
| M2 — Scale, years 2–4 | $1,050,000 | $3,400,000 | $4,450,000 | $118 |
| M3 — Norm, years 4–7 | $3,900,000 | $5,600,000 | $9,500,000 | $70 |
Cost per person certified falls by a factor of six between the pilot and scale, because the curriculum is authored once and the marginal cost of an additional Level 1 learner is a download. The pilot is deliberately expensive per head. That is what a pilot is for, and the figure is published rather than hidden in an average.
A certificate has value. Anything with value can be sold. Transparency International Bangladesh's 2023 household survey found more than half of households paying bribes for public services, and roughly a quarter of households using registered educational institutions reporting corruption.16 Its 2021 education-sector study documented a going rate of Tk 50,000 to Tk 500,000 for an "audit and inspection" of an educational institution.62 The inspection market in Bangladesh already exists and is already for sale. A certification scheme that does not plan for its own certificates being sold is not optimistic; it is negligent.
Countermeasures: assessors are assigned centrally and never to their own district; every audit is countersigned by a second assessor who reviews the evidence pack remotely, including diaspora assessors who have no local relationships at all; the certification decision is taken by a panel that never meets the institution; every certificate number is publicly verifiable; audit findings are spot-audited by an independent reviewer; and a proven sale of a certificate results in permanent debarment of the assessor, publication of the case, and revocation of every certificate that assessor issued in the preceding two years. That last consequence is severe on purpose, because it makes every assessor a monitor of every other.
A policy draft that lists only manageable risks is advocacy. These are ordered by how badly they would damage the proposal, and the first one is a challenge to its central premise.
| Risk | Severity | Response |
|---|---|---|
| The evidence that safeguarding training reduces harm is weak. The 2022 Cochrane review of child-protection training for professionals concluded that the evidence is very uncertain. Certification schemes worldwide have not demonstrated that they make children safer | Fundamental | Stated openly rather than buried. Three consequences follow. First, the Council's theory of change does not rest on training alone but on the combination of trained adults, changed physical arrangements under S5, a functioning complaint route under S2 and S8, and a barred register — the situational and cultural factors the institutional-abuse literature identifies as decisive.3334 Second, the evaluation is pre-registered and empowered to publish a null result. Third, the M4 constitution commits the Council to winding down if ten years produce no measurable effect. A body that cannot say what would falsify it should not be trusted with a child |
| Voluntary certification attracts only the already-safe. This is exactly what happened to England's out-of-school settings accreditation: uptake of four or five settings per authority, and no evidence that it changed parents' behaviour23 | Fundamental | The entire demand-side programme in §10 and the M2 stop condition exist for this. If no board has entered recognition talks and no major donor has made the certificate a condition by year four, the theory has failed and the Council stops calling itself a certifier |
| Voluntary certifiers go out of business. Child Wise, for years Australia's best-known voluntary child-safe certification body, was deregistered in March 2025. Voluntary schemes without a funded demand side do not merely underperform; they close | High | The endowment exists precisely so that the Council is not dependent on certification fee income in its early years, and the cross-subsidy from fee-paying elite institutions is designed to be structural rather than incidental. A certifier that must sell certificates to survive will eventually sell them |
| Sectarian capture. The proposal is framed as a foreign, secular or Islamophobic attack on religious education and dies. India's experience shows how fast this happens46 | High | Ulama veto over religious framing; adoption of ulama-authored proposals as written and credited; the standards published in Bangla first; cadet colleges and elite English-medium boarding schools certified early and at full fee; no position ever taken on curriculum or theology |
| Paper compliance. Institutions produce excellent documentation and change nothing. The Catholic Church's own auditor reports that in one year only 24 of 61 audited dioceses permitted visits to parishes and schools48 | High | Mandatory on-site visits at every cycle; private conversations with children chosen by the assessor; 10% unannounced spot checks drawn centrally; removal from the register for refusing any of the three, published |
| Harm caused by the scheme itself. Fellows or assessors abuse children; a disclosure is mishandled; a falsely accused teacher is destroyed | High | Fellows certified and vetted before any contact, never alone with a child, never sleeping in institutions, always in threes. A published due-process standard for the accused including public clearance where an allegation is disproved — a provision taken directly from Sheikh Ahmadullah's proposal.49 The Council holds itself to S3 and S4 first |
| Certificate fraud. Certificates sold, forged or issued by a captured assessor | High | Central assignment of assessors away from home districts; remote countersignature including by diaspora assessors; decisions by a panel that never meets the institution; publicly verifiable certificate numbers; revocation of an implicated assessor's entire two-year output |
| Government hostility or NGO registration suspension | Medium | The split legal structure means the standards, registers, endowment and appeals sit outside Bangladesh. Operations can be suspended without the institution being destroyed. Rising domestic revenue reduces exposure each year. Note that the foreign-donations regime was liberalised in April 2026, not tightened56 |
| Diaspora factionalism or absentee governance. Overseas boards drift, split along the politics of the country they left, or impose models that do not fit | Medium | Bangladeshi-majority board on a constitutional trigger at year seven; Bangladeshi co-authorship of every standard and module; the Shura Council's veto; trustees who miss three consecutive meetings cease to hold office automatically |
| A safeguarding scandal inside a certified institution | Medium, and inevitable | Plan for it as a certainty rather than a risk. Publish the case, the audit history and what the audit missed, before the press finds it. Independent review of the certification decision. The reputation of a certifier survives failures it discloses and does not survive one it conceals |
| Displacement. Institutions refuse and children move to worse, invisible ones | Medium | Never advocate closure. Registered tier free and easy to enter, so that the first step costs an institution nothing but a self-assessment. Measure the uncertified sector in the evaluation, not just the certified one |
| Cost to poor institutions. A madrasa on donation income cannot afford compliance | Medium | Free below a published income threshold, permanently, funded by cross-subsidy. S5 improvements sequenced to start with the near-zero-cost ones: bed spacing, moving adult sleeping quarters, a lock on a bathroom door, a visitor log. Capital grants for fire safety at the Exemplar tier |
Certificates issued is an activity measure and will be reported as one. It is not evidence that a child is safer, and this document does not pretend otherwise.
| Layer | Measures | Method | Published |
|---|---|---|---|
| Activity | Certificates by level; institutions by tier; audits; unannounced visits; complaints received and resolved | Registry, automatic | Monthly dashboard |
| Compliance | Standard-by-standard scores; corrective actions raised and closed; suspensions; revocations | Audit data | Quarterly, disaggregated |
| Observed practice | Night supervision actually present; adult and child sleeping arrangements; bed spacing; complaint box use; incident logs kept | Unannounced observation on a random sample, by an assessor who did not certify | Annual |
| Children's own accounts | Do you know who to tell; has anyone hurt you; do you feel safe at night; has anything changed | Confidential child survey by the independent evaluator, never by Council staff, in certified and matched uncertified institutions | Annual, in full |
| Impact | Difference in reported and observed harm between certified institutions and matched uncertified controls, over time | Pre-registered quasi-experimental design by a Bangladeshi research institute, with a stepped rollout that creates a legitimate comparison group | Baseline at M0; interim at M2; full at M4 |
The stepped rollout is a deliberate design choice with a cost: it means some institutions wait. It is accepted because a scheme that cannot demonstrate effect will eventually be defunded on suspicion, and because the alternative — certifying everyone at once and asserting success — is how the programmes that burned the cash were run.
Children's own accounts sit deliberately above impact in the hierarchy of what the Council watches, because the Australian Royal Commission's research found that what children identify as making them safe differs systematically from what adults assume, and because a child survey run by the certifier itself would be worthless.33
Every source below was opened and read during the preparation of this draft. Where a figure is an estimate, an advocacy tally or a self-report rather than an official statistic, that is stated in the text at the point of use.
On method. This draft was prepared from primary sources: statutes and official gazettes, national statistical agencies and their data interfaces, published inspection standards, peer-reviewed research, public inquiry reports, and named public statements. Statistical claims were drawn from the World Bank Open Data interface, the UNESCO Institute for Statistics and the UNICEF Data Warehouse directly rather than from secondary summaries. Figures produced by advocacy organisations through media monitoring are identified as such wherever they appear and are never presented as prevalence.
On verification, and its limits. Sources were compiled and checked by the researchers who opened them, and statistical claims were pulled from official data interfaces directly. A planned independent adversarial fact-check of every headline figure and citation did not complete. Three specific cautions follow. The Australian Royal Commission's own domain did not respond during preparation, so its research reports were read from archived copies and their public links should be confirmed before this draft is circulated. Figures on peer abuse in residential care come from different studies measuring different things and must never be added or averaged. And the Ugandan trial did not separately report sexual-violence outcomes, so no claim is made here that the Good School Toolkit reduced sexual violence.
On what is not known. Bangladesh has no current national count of Qawmi madrasas or their students, no count of children living in residential educational institutions, and no setting-disaggregated measurement of violence against children more recent than 2013. The estimates used here are labelled and their ranges are given. Establishing these numbers is itself a milestone-zero task, and the baseline study is budgeted in §10 rather than assumed.
Status. Draft for consultation, version 1.0, September 2026. Circulated for comment to religious scholars, child protection practitioners, university partners, diaspora organisations and prospective funders. Nothing in it has been agreed by any of the bodies named in §8.