The Supervision Week: Debanking Gets Its Receipts, AI Gets Its Rulebook
The OCC's 420,000-page review found the nine largest US banks used standardless reputational-risk policies to cut off customers — with exceptions for the lucrative. Meanwhile the FSB published its 12-practice AI rulebook and Hong Kong moved to human-on-the-loop supervision.
- The OCC released a report on debanking at the nation's nine largest national banks covering 2020 through September 2025; the review was initiated by an August 2025 executive order from President Trump. m:…3509
- The OCC analyzed nearly 100,000 complaints and roughly 420,000 pages of internal documents, finding banks used reputational-risk policies without clear or measurable standards to cut off customers over political, religious, and social views. m:…3509
- Examples include a JPMorgan-owned payments processor closing a merchant's account for selling MAGA gear and a Capital One business line turning away partisan groups on both sides; lucrative clients often received exceptions to industry-wide restrictions, banks have rebranded the policies as franchise or brand risk, and no bank has publicly acknowledged its past debanking decisions. m:…3509
- The Financial Stability Board published its final report on October 15, 2026, titled Sound Practices for Responsible Adoption of Artificial Intelligence (AI), containing 12 sound practices covering organization-wide AI governance and risk management frameworks throughout the AI lifecycle. m:…6309
- In Hong Kong, over 60% of use cases in the HKMA's second GenA.I. Sandbox cohort involved AI agents, with sophisticated outputs achieved on around 30% less computing power than the first cohort; a market-risk agent shortened the cycle of case-specific risk analysis by 95%, as supervisory expectations shift from human-in-the-loop to human-on-the-loop. m:…9973
- The FSB is also consulting on high-level principles for regulatory and supervisory modernization and setting out its work program on AI and stablecoins; separately, the Basel Committee published leverage-ratio assessment reports and the European Banking Authority updated the classification of significant e-money tokens under MiCA. m:…4115
The receipts
For years, complaints that American banks were quietly cutting off disfavored customers were treated as anecdote — the sort of claim that lives in congressional hearings and dies in the footnotes. This week it got a bibliography. The Office of the Comptroller of the Currency released a report on debanking at the nation's nine largest national banks, covering the period from 2020 through September 2025. The review was initiated by an August 2025 executive order from President Trump, and its scale is the point: nearly 100,000 complaints analyzed, roughly 420,000 pages of internal documents reviewed. m:…3509
The findings, as reported by Galaxy Digital Research, describe a system that ran on judgment without yardsticks. Banks used reputational-risk policies without clear or measurable standards to cut off customers over their political, religious, and social views. Industry-wide restrictions on firearms, coal, Arctic drilling, private prisons, tobacco, and digital assets hit small and midsized firms hardest — the customers least able to litigate, lobby, or leave. And the government did not merely watch: Galaxy's reading of the report holds that the crypto industry was targeted by the government in an unfair manner. m:…3509
The case studies give the policy a face. A JPMorgan-owned payments processor closed a merchant's account for selling MAGA gear; a Capital One business line turned away partisan groups on both sides of the aisle. Meanwhile, lucrative clients often received exceptions to the industry-wide restrictions — the rules bent for the profitable. Banks have since rebranded reputational risk as franchise or brand risk, and no bank has publicly acknowledged its past debanking decisions. The vocabulary changed; the accountability did not arrive with it. m:…3509
The rulebooks
While Washington was auditing the past, the standard-setters were writing the future. On October 15, 2026, the Financial Stability Board published its final report, Sound Practices for Responsible Adoption of Artificial Intelligence (AI): 12 sound practices covering organization-wide AI governance and risk management frameworks across the full AI lifecycle. It is the first global template for how a bank is supposed to run a machine that makes decisions, and it lands precisely as banks are handing real decisions to those machines. m:…6309
Hong Kong showed what the template looks like in practice. In the HKMA's second GenA.I. Sandbox cohort, over 60% of use cases involved AI agents — up from a minority in the first — and the cohort achieved sophisticated outputs with around 30% less computing power. The results are concrete enough to be operational: a bank reduced customer-onboarding time by 40% using advanced fraud detection enabled by AI, a wealth-management chatbot achieved a 98% positive-feedback rate, and a market-risk agent shortened the cycle of case-specific risk analysis by 95%. This is no longer experimentation; it is deployment with numbers attached. m:…9973
The supervision is shifting to match. The HKMA's ART framework — Accountability, Resilience, Trust — is now the stated spine of AI adoption in Hong Kong banking, and supervisory expectations are moving from human-in-the-loop to human-on-the-loop oversight for AI applications. The HKMA issued new supervisory guidance on AI adoption aligned with Financial Stability Board standards and is exploring a centralised supervisory registry for AI dependencies in the banking sector — a map, in effect, of every algorithm a bank leans on. m:…9973
Modernization, leverage, and tokens
The weekly regulatory ledger beyond the headlines fills in the picture. The FSB is consulting on high-level principles for regulatory and supervisory modernization — the supervisors modernizing themselves — and setting out its work program regarding AI and stablecoins. The Basel Committee on Banking Supervision published assessment reports on the leverage ratio, and the European Banking Authority updated the classification of significant e-money tokens under MiCA. Taken together, the week's supervisory output has a clear shape: audit what banks did to customers, codify what banks' machines may do, and keep the capital rules honest underneath it all. m:…4115
The irony of the week is worth stating plainly. The same institutions being told by the OCC's report that they acted as enforcers of political and regulatory priorities are now being asked by the FSB and the HKMA to govern autonomous agents acting on their behalf. The question hanging over both halves of the ledger is the same one the OCC report leaves unanswered: when the next debanking happens — and it will be an agent that does it, faster and at greater scale — whose standards will the machine be following, and who will be able to see them? The rulebooks are being written now. The receipts for how they were followed will arrive years later, in another 420,000 pages. m:…3509
What is still unresolved
- Whether any bank faces enforcement action or public remediation from the OCC's findings
- How the FSB's 12 AI practices translate into binding national supervisory rules
- Whether the HKMA's proposed centralised AI-dependency registry becomes mandatory
- How the FSB's AI and stablecoins work program interacts with MiCA's e-money token classification
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.