No AI statute in sight. For banks in North America, model risk guidance now does the work.

Oonagh van den Berg | RAW Compliance

Issue 5 mapped Latin America, where one country has written credit scoring into its AI rules and most of the rest are still debating bills. North America is the reverse. Neither Washington nor Ottawa has an AI statute for banks. What they have is model risk guidance, and in the past six months both sides rewrote it.

The United States rewrote its guidance in April 2026 and took generative and agentic AI out of scope. Canada's OSFI finalised Guideline E-23, which names AI and machine learning, and it applies from 1 May 2027. In the US the harder law is moving to the states, and a federal order is trying to slow that down. A bank cannot read "no federal AI law" as "no AI duties".

United States - the model risk guidance was replaced

On 17 April 2026 the OCC, the Federal Reserve and the FDIC issued revised interagency guidance on model risk management. It rescinds the 2011 guidance (OCC Bulletin 2011-12), the OCC's Model Risk Management handbook booklet, OCC Bulletin 1997-24 on credit scoring models and OCC Bulletin 2021-19 on models supporting BSA/AML compliance. The Federal Reserve issued it as SR 26-2.

Three details matter for AI. First, the agencies say generative AI and agentic AI models are novel and rapidly evolving, and they are not within the scope of the guidance. Second, it is expected to be most relevant to banking organizations with over $30 billion in total assets. Third, the agencies say it sets no enforceable standards, and that non-compliance will not result in supervisory criticism. Banks that used SR 11-7 as the AI validation standard now have a document that says AI of the newest kind is outside it.

Sources: https://www.occ.gov/news-issuances/bulletins/2026/bulletin-2026-13.html ; https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm ; https://www.fdic.gov/model-risk-management-revised-guidance.pdf

United States - the CFPB stepped back, Regulation B did not

In May 2025 the CFPB withdrew 67 guidance documents. They included Consumer Financial Protection Circular 2022-03 and Circular 2023-03, the two circulars on adverse action notices when creditors use complex algorithms or AI. Those circulars are gone as supervisory guidance. The rule underneath them is not. Regulation B, 12 CFR 1002.9, still requires a creditor to give an applicant a statement of specific reasons for an adverse action. A credit model that cannot produce accurate reasons is a Regulation B problem whatever the CFPB says about AI.

Sources: https://www.federalregister.gov/documents/full_text/text/2025/05/12/2025-08286.txt ; https://www.ecfr.gov/current/title-12/chapter-X/part-1002/subpart-A/section-1002.9

United States - the states are writing the AI rules

Colorado is the test case. Its 2024 AI Act, SB 24-205, never took effect. On 14 May 2026 the Governor signed SB 26-189, which repeals and re-enacts it as a narrower automated decision-making technology law. A consequential decision includes decisions about financial or lending services. From 1 January 2027, developers of covered technology must give deployers technical documentation, deployers must give consumers notice at the point of interaction, and after an adverse decision they must give a plain language description of the technology's role. The attorney general enforces it, with a 60-day notice and chance to cure before 1 January 2030. A law firm report says a federal court blocked enforcement of the old act on 27 April 2026 after xAI sued with Department of Justice support. That court order is reported by the firm and was not checked against the docket.

California's privacy regulator has finalised rules on automated decisionmaking technology. The regulations took effect on 1 January 2026, and businesses that use the technology to make significant decisions must comply with the ADMT requirements from 1 January 2027. How far they reach a bank's own customer data depends on existing exemptions, and counsel should confirm that for each product.

Sources: https://leg.colorado.gov/bills/SB26-189 ; https://www.mcdermottlaw.com/insights/colorado-ai-law-in-flux-comprehensive-replacement-bill-signed-after-federal-court-blocks-predecessors-enforcement/ ; https://cppa.ca.gov/announcements/2025/20250923.html

United States - a federal order aimed at the states

Executive Order 14365 of 11 December 2025 sets a policy of a minimal national framework for AI. It directs the Attorney General to set up an AI Litigation Task Force to challenge state AI laws, directs Commerce to evaluate state laws, and ties some federal funding to state positions. An executive order does not repeal a state statute. Until a court or Congress says so, state duties like Colorado's stay on the books, and the Colorado case shows the fight is live.

Source: https://www.federalregister.gov/documents/2025/12/16/2025-23092/ensuring-a-national-policy-framework-for-artificial-intelligence

Canada - OSFI Guideline E-23 names AI

OSFI published the final Guideline E-23 on Model Risk Management on 11 September 2025. It takes effect for all federally regulated financial institutions on 1 May 2027. OSFI describes it as covering traditional models and emerging technologies such as AI and machine learning. It kept the definition of a model deliberately broad, and says only models that carry risk to the institution need to be on the model inventory. OSFI moved the date from a proposed twelve months to 1 May 2027 after industry asked for longer. This is the closest thing in North America to a bank-specific AI model rule, and it has a date.

Sources: https://www.osfi-bsif.gc.ca/en/guidance/guidance-library/guideline-e-23-model-risk-management-2027-letter ; https://www.osfi-bsif.gc.ca/en/news/backgrounder-guideline-e-23-model-risk-management

Canada - OSFI on generative, agentic and frontier AI

OSFI's July 2026 technology risk bulletin on generative and agentic AI sets out sound practices tied to Guidelines B-13 (technology and cyber risk), E-21 (operational risk and resilience) and B-10 (third-party risk). It covers governance, limits on autonomy, human oversight, data lineage and third-party dependencies. It is a bulletin, not a new guideline, so existing guidelines are the hook. On 8 October 2026 OSFI's semi-annual risk outlook focused on frontier AI amplifying cyber, technology, third-party and reputational risk, and said boards and senior management remain accountable. The second Financial Industry Forum on AI (FIFAI II) report followed in March 2026 and stresses that institutions stay accountable for AI-driven consumer outcomes.

Sources: https://www.osfi-bsif.gc.ca/en/risks/technology-cyber-risk-management/technology-risk-bulletin/generative-agentic-artificial-intelligence-implications-technology-cyber-security-operational ; https://www.osfi-bsif.gc.ca/en/news/osfi-releases-fall-risk-update-focused-frontier-ai ; https://www.canada.ca/en/financial-consumer-agency/news/2026/03/final-report-from-the-second-financial-industry-forum-on-artificial-intelligence-fifai-2-now-available.html

Canada - no AI statute, a closed transparency consultation, and Quebec

Canada's proposed Artificial Intelligence and Data Act was part of Bill C-27, which law firm commentary reports died when Parliament was prorogued in January 2025. Between 23 July and 23 September 2026 the federal government consulted on AI transparency: identifying AI-generated content, telling people when they are dealing with an AI system, information about AI systems, tracking serious incidents and tracking AI agents. The consultation is closed and a What We Heard report is still to come. No binding AI transparency rule has been announced.

Quebec is already binding. Section 12.1 of the private sector privacy act, in force since 22 September 2023, requires a business that uses personal information to make a decision based exclusively on automated processing to tell the person no later than when it informs them of the decision, and to give more detail on request.

Sources: https://ised-isde.canada.ca/site/ised/en/have-your-say-advancing-ai-transparency-canada ; https://www.legisquebec.gouv.qc.ca/fr/version/lc/p-39.1?code=se%3A12_1 ; https://canada-insights.bakermckenzie.com/2025/04/17/an-election-is-looming-the-future-of-canadian-ai-legislation/

Why this matters beyond North America

The pattern is the same one the series keeps finding. Regulators do not need an AI law to hold a bank to account for an AI model. They use model risk, consumer credit, privacy and operational resilience rules that already exist. What changes is the paperwork. The US guidance tells you what it does not cover. Canada tells you what it will expect from May 2027.

What bank boards should do now

1. Decide which standard validates your generative and agentic AI. The US guidance leaves it out, and OSFI's bulletin points to B-13, E-21 and B-10.

2. Plan the Canadian model inventory now. E-23 applies from 1 May 2027 to every model that carries risk, AI and machine learning included.

3. Keep adverse action reasons accurate. Regulation B still applies to credit models, with or without CFPB circulars.

4. Map state exposure by product. Colorado's duties start on 1 January 2027 and California's ADMT requirements on the same date.

5. Do not treat the federal order as a pause on state law. Watch the courts.

All issues: https://www.rawcompliance.ai/blog/ai-regulatory-market-map

Source and status note

Positions checked against official materials available on 11 October 2026: OCC, Federal Reserve and FDIC for the model risk guidance, the Federal Register for the CFPB withdrawal and Executive Order 14365, the Colorado General Assembly, the California Privacy Protection Agency, OSFI, ISED and the Financial Consumer Agency of Canada, and Legisquebec. The federal court order on Colorado's earlier act, and the lapse of Bill C-27, are reported by law firms and were not checked against court or parliamentary records. Regulation B is cited from the current eCFR. This is general information, not legal advice.

THE AI REGULATION MARKET MAP • NORTH AMERICA • RAW COMPLIANCE

No AI statute in sight. For banks in North America, model risk guidance now does the work.

Oonagh van den Berg | RAW Compliance

Issue 5 mapped Latin America, where one country has written credit scoring into its AI rules and most of the rest are still debating bills. North America is the reverse. Neither Washington nor Ottawa has an AI statute for banks. What they have is model risk guidance, and in the past six months both sides rewrote it.

The United States rewrote its guidance in April 2026 and took generative and agentic AI out of scope. Canada's OSFI finalised Guideline E-23, which names AI and machine learning, and it applies from 1 May 2027. In the US the harder law is moving to the states, and a federal order is trying to slow that down. A bank cannot read "no federal AI law" as "no AI duties".

United States - the model risk guidance was replaced

On 17 April 2026 the OCC, the Federal Reserve and the FDIC issued revised interagency guidance on model risk management. It rescinds the 2011 guidance (OCC Bulletin 2011-12), the OCC's Model Risk Management handbook booklet, OCC Bulletin 1997-24 on credit scoring models and OCC Bulletin 2021-19 on models supporting BSA/AML compliance. The Federal Reserve issued it as SR 26-2.

Three details matter for AI. First, the agencies say generative AI and agentic AI models are novel and rapidly evolving, and they are not within the scope of the guidance. Second, it is expected to be most relevant to banking organizations with over $30 billion in total assets. Third, the agencies say it sets no enforceable standards, and that non-compliance will not result in supervisory criticism. Banks that used SR 11-7 as the AI validation standard now have a document that says AI of the newest kind is outside it.

Sources: https://www.occ.gov/news-issuances/bulletins/2026/bulletin-2026-13.html ; https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm ; https://www.fdic.gov/model-risk-management-revised-guidance.pdf

United States - the CFPB stepped back, Regulation B did not

In May 2025 the CFPB withdrew 67 guidance documents. They included Consumer Financial Protection Circular 2022-03 and Circular 2023-03, the two circulars on adverse action notices when creditors use complex algorithms or AI. Those circulars are gone as supervisory guidance. The rule underneath them is not. Regulation B, 12 CFR 1002.9, still requires a creditor to give an applicant a statement of specific reasons for an adverse action. A credit model that cannot produce accurate reasons is a Regulation B problem whatever the CFPB says about AI.

Sources: https://www.federalregister.gov/documents/full_text/text/2025/05/12/2025-08286.txt ; https://www.ecfr.gov/current/title-12/chapter-X/part-1002/subpart-A/section-1002.9

United States - the states are writing the AI rules

Colorado is the test case. Its 2024 AI Act, SB 24-205, never took effect. On 14 May 2026 the Governor signed SB 26-189, which repeals and re-enacts it as a narrower automated decision-making technology law. A consequential decision includes decisions about financial or lending services. From 1 January 2027, developers of covered technology must give deployers technical documentation, deployers must give consumers notice at the point of interaction, and after an adverse decision they must give a plain language description of the technology's role. The attorney general enforces it, with a 60-day notice and chance to cure before 1 January 2030. A law firm report says a federal court blocked enforcement of the old act on 27 April 2026 after xAI sued with Department of Justice support. That court order is reported by the firm and was not checked against the docket.

California's privacy regulator has finalised rules on automated decisionmaking technology. The regulations took effect on 1 January 2026, and businesses that use the technology to make significant decisions must comply with the ADMT requirements from 1 January 2027. How far they reach a bank's own customer data depends on existing exemptions, and counsel should confirm that for each product.

Sources: https://leg.colorado.gov/bills/SB26-189 ; https://www.mcdermottlaw.com/insights/colorado-ai-law-in-flux-comprehensive-replacement-bill-signed-after-federal-court-blocks-predecessors-enforcement/ ; https://cppa.ca.gov/announcements/2025/20250923.html

United States - a federal order aimed at the states

Executive Order 14365 of 11 December 2025 sets a policy of a minimal national framework for AI. It directs the Attorney General to set up an AI Litigation Task Force to challenge state AI laws, directs Commerce to evaluate state laws, and ties some federal funding to state positions. An executive order does not repeal a state statute. Until a court or Congress says so, state duties like Colorado's stay on the books, and the Colorado case shows the fight is live.

Source: https://www.federalregister.gov/documents/2025/12/16/2025-23092/ensuring-a-national-policy-framework-for-artificial-intelligence

Canada - OSFI Guideline E-23 names AI

OSFI published the final Guideline E-23 on Model Risk Management on 11 September 2025. It takes effect for all federally regulated financial institutions on 1 May 2027. OSFI describes it as covering traditional models and emerging technologies such as AI and machine learning. It kept the definition of a model deliberately broad, and says only models that carry risk to the institution need to be on the model inventory. OSFI moved the date from a proposed twelve months to 1 May 2027 after industry asked for longer. This is the closest thing in North America to a bank-specific AI model rule, and it has a date.

Sources: https://www.osfi-bsif.gc.ca/en/guidance/guidance-library/guideline-e-23-model-risk-management-2027-letter ; https://www.osfi-bsif.gc.ca/en/news/backgrounder-guideline-e-23-model-risk-management

Canada - OSFI on generative, agentic and frontier AI

OSFI's July 2026 technology risk bulletin on generative and agentic AI sets out sound practices tied to Guidelines B-13 (technology and cyber risk), E-21 (operational risk and resilience) and B-10 (third-party risk). It covers governance, limits on autonomy, human oversight, data lineage and third-party dependencies. It is a bulletin, not a new guideline, so existing guidelines are the hook. On 8 October 2026 OSFI's semi-annual risk outlook focused on frontier AI amplifying cyber, technology, third-party and reputational risk, and said boards and senior management remain accountable. The second Financial Industry Forum on AI (FIFAI II) report followed in March 2026 and stresses that institutions stay accountable for AI-driven consumer outcomes.

Sources: https://www.osfi-bsif.gc.ca/en/risks/technology-cyber-risk-management/technology-risk-bulletin/generative-agentic-artificial-intelligence-implications-technology-cyber-security-operational ; https://www.osfi-bsif.gc.ca/en/news/osfi-releases-fall-risk-update-focused-frontier-ai ; https://www.canada.ca/en/financial-consumer-agency/news/2026/03/final-report-from-the-second-financial-industry-forum-on-artificial-intelligence-fifai-2-now-available.html

Canada - no AI statute, a closed transparency consultation, and Quebec

Canada's proposed Artificial Intelligence and Data Act was part of Bill C-27, which law firm commentary reports died when Parliament was prorogued in January 2025. Between 23 July and 23 September 2026 the federal government consulted on AI transparency: identifying AI-generated content, telling people when they are dealing with an AI system, information about AI systems, tracking serious incidents and tracking AI agents. The consultation is closed and a What We Heard report is still to come. No binding AI transparency rule has been announced.

Quebec is already binding. Section 12.1 of the private sector privacy act, in force since 22 September 2023, requires a business that uses personal information to make a decision based exclusively on automated processing to tell the person no later than when it informs them of the decision, and to give more detail on request.

Sources: https://ised-isde.canada.ca/site/ised/en/have-your-say-advancing-ai-transparency-canada ; https://www.legisquebec.gouv.qc.ca/fr/version/lc/p-39.1?code=se%3A12_1 ; https://canada-insights.bakermckenzie.com/2025/04/17/an-election-is-looming-the-future-of-canadian-ai-legislation/

Why this matters beyond North America

The pattern is the same one the series keeps finding. Regulators do not need an AI law to hold a bank to account for an AI model. They use model risk, consumer credit, privacy and operational resilience rules that already exist. What changes is the paperwork. The US guidance tells you what it does not cover. Canada tells you what it will expect from May 2027.

What bank boards should do now

1. Decide which standard validates your generative and agentic AI. The US guidance leaves it out, and OSFI's bulletin points to B-13, E-21 and B-10.

2. Plan the Canadian model inventory now. E-23 applies from 1 May 2027 to every model that carries risk, AI and machine learning included.

3. Keep adverse action reasons accurate. Regulation B still applies to credit models, with or without CFPB circulars.

4. Map state exposure by product. Colorado's duties start on 1 January 2027 and California's ADMT requirements on the same date.

5. Do not treat the federal order as a pause on state law. Watch the courts.

All issues: https://www.rawcompliance.ai/blog/ai-regulatory-market-map

Source and status note

Positions checked against official materials available on 11 October 2026: OCC, Federal Reserve and FDIC for the model risk guidance, the Federal Register for the CFPB withdrawal and Executive Order 14365, the Colorado General Assembly, the California Privacy Protection Agency, OSFI, ISED and the Financial Consumer Agency of Canada, and Legisquebec. The federal court order on Colorado's earlier act, and the lapse of Bill C-27, are reported by law firms and were not checked against court or parliamentary records. Regulation B is cited from the current eCFR. This is general information, not legal advice.

THE AI REGULATION MARKET MAP • NORTH AMERICA • RAW COMPLIANCE

No AI statute in sight. For banks in North America, model risk guidance now does the work.

Oonagh van den Berg | RAW Compliance

Issue 5 mapped Latin America, where one country has written credit scoring into its AI rules and most of the rest are still debating bills. North America is the reverse. Neither Washington nor Ottawa has an AI statute for banks. What they have is model risk guidance, and in the past six months both sides rewrote it.

The United States rewrote its guidance in April 2026 and took generative and agentic AI out of scope. Canada's OSFI finalised Guideline E-23, which names AI and machine learning, and it applies from 1 May 2027. In the US the harder law is moving to the states, and a federal order is trying to slow that down. A bank cannot read "no federal AI law" as "no AI duties".

United States - the model risk guidance was replaced

On 17 April 2026 the OCC, the Federal Reserve and the FDIC issued revised interagency guidance on model risk management. It rescinds the 2011 guidance (OCC Bulletin 2011-12), the OCC's Model Risk Management handbook booklet, OCC Bulletin 1997-24 on credit scoring models and OCC Bulletin 2021-19 on models supporting BSA/AML compliance. The Federal Reserve issued it as SR 26-2.

Three details matter for AI. First, the agencies say generative AI and agentic AI models are novel and rapidly evolving, and they are not within the scope of the guidance. Second, it is expected to be most relevant to banking organizations with over $30 billion in total assets. Third, the agencies say it sets no enforceable standards, and that non-compliance will not result in supervisory criticism. Banks that used SR 11-7 as the AI validation standard now have a document that says AI of the newest kind is outside it.

Sources: https://www.occ.gov/news-issuances/bulletins/2026/bulletin-2026-13.html ; https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm ; https://www.fdic.gov/model-risk-management-revised-guidance.pdf

United States - the CFPB stepped back, Regulation B did not

In May 2025 the CFPB withdrew 67 guidance documents. They included Consumer Financial Protection Circular 2022-03 and Circular 2023-03, the two circulars on adverse action notices when creditors use complex algorithms or AI. Those circulars are gone as supervisory guidance. The rule underneath them is not. Regulation B, 12 CFR 1002.9, still requires a creditor to give an applicant a statement of specific reasons for an adverse action. A credit model that cannot produce accurate reasons is a Regulation B problem whatever the CFPB says about AI.

Sources: https://www.federalregister.gov/documents/full_text/text/2025/05/12/2025-08286.txt ; https://www.ecfr.gov/current/title-12/chapter-X/part-1002/subpart-A/section-1002.9

United States - the states are writing the AI rules

Colorado is the test case. Its 2024 AI Act, SB 24-205, never took effect. On 14 May 2026 the Governor signed SB 26-189, which repeals and re-enacts it as a narrower automated decision-making technology law. A consequential decision includes decisions about financial or lending services. From 1 January 2027, developers of covered technology must give deployers technical documentation, deployers must give consumers notice at the point of interaction, and after an adverse decision they must give a plain language description of the technology's role. The attorney general enforces it, with a 60-day notice and chance to cure before 1 January 2030. A law firm report says a federal court blocked enforcement of the old act on 27 April 2026 after xAI sued with Department of Justice support. That court order is reported by the firm and was not checked against the docket.

California's privacy regulator has finalised rules on automated decisionmaking technology. The regulations took effect on 1 January 2026, and businesses that use the technology to make significant decisions must comply with the ADMT requirements from 1 January 2027. How far they reach a bank's own customer data depends on existing exemptions, and counsel should confirm that for each product.

Sources: https://leg.colorado.gov/bills/SB26-189 ; https://www.mcdermottlaw.com/insights/colorado-ai-law-in-flux-comprehensive-replacement-bill-signed-after-federal-court-blocks-predecessors-enforcement/ ; https://cppa.ca.gov/announcements/2025/20250923.html

United States - a federal order aimed at the states

Executive Order 14365 of 11 December 2025 sets a policy of a minimal national framework for AI. It directs the Attorney General to set up an AI Litigation Task Force to challenge state AI laws, directs Commerce to evaluate state laws, and ties some federal funding to state positions. An executive order does not repeal a state statute. Until a court or Congress says so, state duties like Colorado's stay on the books, and the Colorado case shows the fight is live.

Source: https://www.federalregister.gov/documents/2025/12/16/2025-23092/ensuring-a-national-policy-framework-for-artificial-intelligence

Canada - OSFI Guideline E-23 names AI

OSFI published the final Guideline E-23 on Model Risk Management on 11 September 2025. It takes effect for all federally regulated financial institutions on 1 May 2027. OSFI describes it as covering traditional models and emerging technologies such as AI and machine learning. It kept the definition of a model deliberately broad, and says only models that carry risk to the institution need to be on the model inventory. OSFI moved the date from a proposed twelve months to 1 May 2027 after industry asked for longer. This is the closest thing in North America to a bank-specific AI model rule, and it has a date.

Sources: https://www.osfi-bsif.gc.ca/en/guidance/guidance-library/guideline-e-23-model-risk-management-2027-letter ; https://www.osfi-bsif.gc.ca/en/news/backgrounder-guideline-e-23-model-risk-management

Canada - OSFI on generative, agentic and frontier AI

OSFI's July 2026 technology risk bulletin on generative and agentic AI sets out sound practices tied to Guidelines B-13 (technology and cyber risk), E-21 (operational risk and resilience) and B-10 (third-party risk). It covers governance, limits on autonomy, human oversight, data lineage and third-party dependencies. It is a bulletin, not a new guideline, so existing guidelines are the hook. On 8 October 2026 OSFI's semi-annual risk outlook focused on frontier AI amplifying cyber, technology, third-party and reputational risk, and said boards and senior management remain accountable. The second Financial Industry Forum on AI (FIFAI II) report followed in March 2026 and stresses that institutions stay accountable for AI-driven consumer outcomes.

Sources: https://www.osfi-bsif.gc.ca/en/risks/technology-cyber-risk-management/technology-risk-bulletin/generative-agentic-artificial-intelligence-implications-technology-cyber-security-operational ; https://www.osfi-bsif.gc.ca/en/news/osfi-releases-fall-risk-update-focused-frontier-ai ; https://www.canada.ca/en/financial-consumer-agency/news/2026/03/final-report-from-the-second-financial-industry-forum-on-artificial-intelligence-fifai-2-now-available.html

Canada - no AI statute, a closed transparency consultation, and Quebec

Canada's proposed Artificial Intelligence and Data Act was part of Bill C-27, which law firm commentary reports died when Parliament was prorogued in January 2025. Between 23 July and 23 September 2026 the federal government consulted on AI transparency: identifying AI-generated content, telling people when they are dealing with an AI system, information about AI systems, tracking serious incidents and tracking AI agents. The consultation is closed and a What We Heard report is still to come. No binding AI transparency rule has been announced.

Quebec is already binding. Section 12.1 of the private sector privacy act, in force since 22 September 2023, requires a business that uses personal information to make a decision based exclusively on automated processing to tell the person no later than when it informs them of the decision, and to give more detail on request.

Sources: https://ised-isde.canada.ca/site/ised/en/have-your-say-advancing-ai-transparency-canada ; https://www.legisquebec.gouv.qc.ca/fr/version/lc/p-39.1?code=se%3A12_1 ; https://canada-insights.bakermckenzie.com/2025/04/17/an-election-is-looming-the-future-of-canadian-ai-legislation/

Why this matters beyond North America

The pattern is the same one the series keeps finding. Regulators do not need an AI law to hold a bank to account for an AI model. They use model risk, consumer credit, privacy and operational resilience rules that already exist. What changes is the paperwork. The US guidance tells you what it does not cover. Canada tells you what it will expect from May 2027.

What bank boards should do now

1. Decide which standard validates your generative and agentic AI. The US guidance leaves it out, and OSFI's bulletin points to B-13, E-21 and B-10.

2. Plan the Canadian model inventory now. E-23 applies from 1 May 2027 to every model that carries risk, AI and machine learning included.

3. Keep adverse action reasons accurate. Regulation B still applies to credit models, with or without CFPB circulars.

4. Map state exposure by product. Colorado's duties start on 1 January 2027 and California's ADMT requirements on the same date.

5. Do not treat the federal order as a pause on state law. Watch the courts.

All issues: https://www.rawcompliance.ai/blog/ai-regulatory-market-map

Source and status note

Positions checked against official materials available on 11 October 2026: OCC, Federal Reserve and FDIC for the model risk guidance, the Federal Register for the CFPB withdrawal and Executive Order 14365, the Colorado General Assembly, the California Privacy Protection Agency, OSFI, ISED and the Financial Consumer Agency of Canada, and Legisquebec. The federal court order on Colorado's earlier act, and the lapse of Bill C-27, are reported by law firms and were not checked against court or parliamentary records. Regulation B is cited from the current eCFR. This is general information, not legal advice.

THE AI REGULATION MARKET MAP • NORTH AMERICA • RAW COMPLIANCE