AI Regulation Tracker / Reference
AI Laws by Profession
The AI rules that change what you must do, organized by profession and jurisdiction. We update this as new rules land.
Lawyers and litigators
Courts and bars across jurisdictions now treat verifying every AI-generated citation as a professional duty, with sanctions, fines and referrals for those who file fabricated authority.
- US Federal The Fifth Circuit sanctioned an AI-fabricated brief as more circuits converge.
- Colombia The Supreme Court annulled a ruling built on nonexistent case citations.
- Brazil The TST labor court fined a lawyer and company for AI-invented precedents.
- Brazil The Sao Paulo bar makes partners answerable for associates' AI use.
- Canada Law societies and Legal Aid Ontario attach AI competence and attestation duties.
- Qatar The QICDRC requires lawyers to identify, verify and disclose AI in filings.
- Australia The Fair Work Commission drafts a duty to disclose and verify AI-prepared filings.
- Japan The bar sets five rules for lawyers using generative AI.
Physicians and clinicians
Medical councils and device regulators keep the clinician responsible for AI-assisted care and hold AI medical software to the full device rules.
- Brazil CFM Resolution 2.454 makes physicians own AI-assisted decisions.
- Canada The CPSO keeps Ontario doctors accountable for AI in clinical care.
- New Zealand The Medical Council keeps doctors responsible for every AI-assisted decision.
- Australia The TGA applies full Software as a Medical Device rules to AI.
- Colorado, US HB 26-1195 restricts AI in therapeutic communication without oversight.
- Taiwan The TFDA sets validation criteria for diagnostic AI devices.
- South Korea The MFDS opens an approval pathway for generative-AI medical devices.
- Canada Health Canada sets change-control expectations for ML medical devices.
Financial advisers and broker-dealers
Financial regulators worldwide hold the licensee responsible for AI-assisted advice and expect a named human to own AI-driven decisions.
- US Federal FINRA makes generative-AI supervision a 2026 exam priority.
- US Federal The SEC withdrew its AI rule but AI-washing enforcement continues.
- Australia ASIC holds advisers responsible for any AI-assisted recommendation.
- New Zealand The FMA says conduct duties apply to AI in advice, with a review planned.
- UAE The Central Bank sets responsible-AI duties for licensed finance firms.
- Taiwan The FSC sets lifecycle AI governance for banks, insurers and securities.
- Japan The FSA keeps the suitability and explanation duty on AI advice.
- South Korea A draft financial-AI code would put a named human on every AI credit call.
Insurers and benefit managers
Insurance regulators bar AI-only coverage denials, require reporting of AI-aided decisions, and insurers themselves are writing AI out of coverage.
- Washington, US SB 5395 bars AI as the sole basis to deny prior authorization.
- Maryland, US HB 1563 makes insurers report whether AI drove each denial.
- Iowa, US HF 2635 bars AI-only prior-auth denials and requires a same-specialty reviewer.
- Global New ISO endorsements let carriers exclude generative-AI losses.
- Taiwan Insurer self-regulation sets explainability duties on underwriting and claims AI.
- Brazil SUSEP's permanent sandbox is the supervised route for AI in claims.
HR, employers and hiring
Employment and privacy regulators add notice, opt-out, human-review and workplace-safety duties to AI used in hiring and management.
- California, US CCPA ADMT rules add notice and opt-out duties for AI in hiring.
- US Federal A DOJ opinion attacks the EEOC disparate-impact rule; exposure survives.
- NSW, Australia A work-health duty governs AI work allocation and algorithmic management.
- Taiwan Draft rules require notice, a human decision and an explainable hiring result.
- South Korea AI hiring is treated as high-impact, with disclose-and-justify duties.
Mental-health and companion-AI providers
A fast-growing set of laws bars AI from posing as a therapist and imposes disclosure, crisis-protocol and minor-safety duties on companion chatbots, often with a private right of action.
- Illinois, US The WOPR Act bars AI from delivering therapy.
- Tennessee, US SB 1580 bars AI from posing as a mental health professional.
- California, US SB 243 sets companion-chatbot safety duties with a private right of action.
- Oregon, US SB 1546 adds companion-AI duties backed by a private right of action.
- Nebraska, US The Conversational AI Safety Act sets disclosure and crisis duties.
- China Anthropomorphic-AI measures require disclosure and safeguards for companion bots.
Marketers, platforms and content
Deepfake, impersonation and content-labeling rules create takedown, disclosure and screening duties for anyone producing or distributing synthetic media.
- New Jersey, US Deceptive AI media for an unlawful purpose is a third-degree crime.
- Peru Law 32314 makes AI deepfakes an aggravating criminal factor.
- US Federal The FTC impersonation rule reaches AI voice cloning.
- European Union Article 5 bans AI that generates non-consensual intimate imagery.
- Australia The eSafety codes make AI nudify tools block children or face penalties.
- United Kingdom A ruling put AI output on the hook for reproducing trademarks.
- Turkiye A bill would make platforms answer for AI-generated content.
Developers, AI vendors and copyright
Training-data litigation, copyright policy and vendor terms are setting hard duties around what firms may train on and what they inherit from vendors.
- US Federal Bartz v. Anthropic set a $1.5 billion benchmark for AI copyright exposure.
- US Federal The Copyright Office and courts hold that AI-alone work is unregistrable.
- Australia Australia ruled out a text-and-data-mining copyright exception.
- Japan Article 30-4's free-to-train rule has real carve-outs for rights-holders.
- US Federal The NO FAKES Act would create a federal digital-replica right.
- Global Atlassian moves to train its AI on customer data; review and opt out.
- Global Most enterprise SaaS contracts claim AI-training rights; red-line them.
Compliance, security and governance leaders
Standards bodies, security researchers and prudential regulators are converging on a documented AI governance, model-risk and attack-surface duty.
- Global OWASP frames prompt injection as a structural flaw in agentic AI.
- Global A NIST crosswalk maps the AI RMF to the certifiable ISO/IEC 42001.
- US Federal NIST is building an AI RMF profile for critical infrastructure.
- Australia APRA CPS 230 treats AI and cloud vendors as material service providers.
- UAE The DIFC now enforces Regulation 10 on autonomous systems.
- Canada OSFI E-23 sets model-risk expectations for AI and ML.
- Global The first reported AI-run ransomware maps to existing breach duties.
Accountants and tax practitioners
Professional bodies and revenue authorities keep the practitioner responsible for verifying AI output in regulated work.
- US Federal The IRS applies Circular 230 duties to AI in tax practice.
- Canada CPA Ontario keeps accountants answerable for AI-assisted work.
- Japan Patent attorneys are warned that pasting an invention into AI risks confidentiality.
Sponsored Training
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.