Mar 20, 2025
National Policy Framework for AI
White House blueprint: a single preemptive federal framework, age-assurance for minors, courts to lead on training-data copyright, and no new AI-specific agency.
Policy blueprint
Praxis Lab ◆ Regulatory Intelligence
Live · Jan–Aug 2026
The US AI rulebook is being written in real time.
A live map of the American AI-governance landscape — federal orders, agency guidance, and the state statutes racing ahead of Congress. Toggle between when each measure was signed and when it actually bites. Filter by state, instrument, or subject; every card opens to the operative details.
34
Measures tracked
15
Jurisdictions
10
State laws enacted in 2026
8
In force right now
The 2026 landscape, in five lines
The states are writing the operative rulebook. With no comprehensive federal AI statute, fourteen states plus the federal executive account for all 34 tracked measures — and ten new state laws were enacted in the first half of 2026 alone.
Companion-chatbot law is the year's defining wave. California, Washington, Oregon, Idaho, Nebraska, Iowa, and Connecticut now require non-human disclosure, self-harm detection with crisis referral, and heightened protection of minors — on a shared statutory template with meaningful state-by-state divergence.
Washington is pushing back on the states. Executive Order 14365 directs agencies to identify and challenge state AI laws, and the TRUMP AMERICA AI Act discussion draft would codify partial federal preemption — the central structural fight of 2026.
Frontier-model regimes are converging on California's TFAIA. New York's S.8828 overhaul realigned the RAISE Act to the California model — incident reporting, catastrophic-risk summaries, whistleblower protection — days after the original took effect.
Colorado shows the ceiling. The most ambitious state AI act of 2024 (S.B. 205) stalled in litigation and was replaced by the narrower S.B. 189 ADMT law — a template for how broad duty-of-care regimes get trimmed to consumer rights.
Compliance countdown
Next effective dates, counted live against the day you open this page. From July 18, 2026 the queue runs: California procurement reports (Jul 28, 2026) → California AI Transparency Act (Aug 2, 2026) → Connecticut S.B. 5 & H.B. 5222 (Oct 1, 2026) → Colorado S.B. 189 and Oregon S.B. 1546 (Jan 1, 2027) → Washington H.B. 1170 & H.B. 2225 (Feb 1, 2027) → Idaho S.B. 1297 and Iowa S.F. 2417 (Jul 1, 2027).
10
days until effect
◆ Jul 28, 2026
Calif.orders AI safety certifications for state procurement
Calif.AI-procurement recommendations come due
15
days until effect
◆ Aug 2, 2026
Calif.AI Transparency Act takes effect
75
days until effect
◆ Oct 1, 2026
Conn.enacts a sweeping online-safety & AI law
Conn.adds GenAI subscription-notice rules
167
days until effect
◆ Jan 1, 2027
Ore.enacts an AI-companion law
Colo.replaces its 2024 AI Act with a narrower ADMT law
198
days until effect
◆ Feb 1, 2027
Wash.enacts an AI content-provenance law
Wash.enacts an AI-companion chatbot law
348
days until effect
◆ Jul 1, 2027
Idahoenacts the Conversational AI Safety Act
Iowaenacts a conversational-AI law
Where the action is
tap a lit state to filter the timeline →
Federal backdrop & bills in motion
Dec 11, 2025
Executive Order 14365
Directs a multi-agency effort to identify and challenge state AI laws, and tasks the FCC with weighing a preemptive federal disclosure standard.
Preemption pushMar 18, 2026
TRUMP AMERICA AI Act (draft)
Sen. Blackburn’s 291-page discussion draft codifying EO 14365 — partial preemption on minors, developer/deployer liability, and digital replicas.
Discussion draftMay 21, 2025
DEFIANCE Act (S. 1837)
Would create a federal civil remedy for nonconsensual intimate deepfakes, with statutory and punitive damages plus fees.
Pending billThe 2026 timeline
30 shown
Filter by instrument
Filter by subject
Showing
Jan2026
5 items
Takes effect
Ill.
In force
Jan 1, 2026
Illinois civil-rights AI amendments take effect
AI-driven actions are folded into the state civil-rights act, reaching employment discrimination and digital replicas.
H.B. 3773 (2024)▸ details
◆ Effective Jan 1, 2026
- Employers become answerable for discriminatory automated decisions.
- Ties into the state’s digital-replica and likeness protections.
Takes effect
Tex.
In force
Jan 1, 2026
Texas Responsible AI Governance Act takes effect
A framework for AI development and use, including a sandbox that relaxes certain legal constraints for testing.
TRAIGA (2025)▸ details
◆ Effective Jan 1, 2026
- Creates a state regulatory system for AI development and deployment.
- Sandbox program lets developers test with reduced legal exposure.
Takes effect
Calif.
In force
Jan 1, 2026
California frontier training-data disclosures begin
Frontier developers must publish training-data documentation for GenAI released or materially changed after 2026.
TFAIA (2025)▸ details
◆ Effective Jan 1, 2026
- Part of the Transparency in Frontier AI Act, aimed at catastrophic risk.
- Adds whistleblower protection for employees flagging critical safety issues.
Takes effect
Calif.
In force
Jan 1, 2026
California chatbot guardrails (S.B. 243) take effect
Disclosure and safety guardrails for consumer-facing chatbots become enforceable.
Cal. Bus. & Prof. §22601▸ details
◆ Effective Jan 1, 2026
- Requires clear disclosure that a user is interacting with AI.
- Layers onto California’s broader consumer-protection regime.
Guidance
Mich.
Jan 14, 2026
Michigan warns insurers: AI decisions still answer to old law
State regulator reminds financial-services firms that automated decisions remain fully subject to existing statutes.
MDIFS Bulletin 2026-03▸ details
- Applies across insurance, banking and credit-union supervision.
- Endorses Treasury’s 2024 “AI in Financial Services” report as a reference.
- No new rule — a statement that AI is not a compliance loophole.
Feb2026
2 items
Federal guidance
U.S.
Feb 19, 2026
Treasury issues an AI lexicon and a financial-sector risk framework
Two companion documents: a shared AI vocabulary and a financial-services risk-management framework built on the NIST RMF.
Treasury: FS AI RMF + AI Lexicon▸ details
- The Financial Services AI RMF adapts NIST’s framework to banking and markets.
- The Lexicon aims to standardize terminology across regulated firms.
Guidance
Conn.
Feb 25, 2026
Connecticut AG maps existing law onto AI
The attorney general spells out how current statutes already reach AI development and use.
Conn. AG memorandum▸ details
- Covers civil-rights & nondiscrimination law.
- Covers data privacy and breach obligations.
- Covers antitrust and consumer protection.
Mar2026
6 items
Takes effect
N.Y.
In force
Mar 19, 2026
New York RAISE Act hits its effective date
Reporting and safety duties on large frontier developers take effect — days before the legislature reworks the whole regime.
RAISE Act (2025)▸ details
◆ Effective Mar 19, 2026
- Original RAISE Act imposed catastrophic-risk safeguards on advanced-AI developers.
- Superseded almost immediately by the S.8828 overhaul (Mar 27).
Signed into law
Wash.
Effective Feb 1 ’27
Mar 24, 2026
Washington enacts an AI content-provenance law
Large GenAI providers must embed tamper-resistant watermarks or metadata so AI content’s origin can be traced.
H.B. 1170▸ details
◆ Effective Feb 1, 2027
- “Covered providers” = GenAI systems with 1M+ monthly Washington users.
- Provenance data must be difficult to remove or tamper with.
- Effective February 1, 2027.
Signed into law
Wash.
Effective Feb 1 ’27
Mar 24, 2026
Washington enacts an AI-companion chatbot law
Companion bots must disclose they aren’t human, run self-harm detection with crisis referrals, and add stricter under-18 protections.
H.B. 2225▸ details
◆ Effective Feb 1, 2027
- Non-human disclosure at the start and every three hours; bots may never claim to be human.
- For known minors: hourly notice, no sexual content, no manipulative bonding tactics.
- Enforced as a state consumer-protection violation; effective Feb 1, 2027.
Signed into law
N.Y.
Mar 27, 2026
New York overhauls the RAISE Act
Realigns New York with California’s TFAIA — moving enforcement to the Dept. of Financial Services and easing penalties.
S. 8828▸ details
- 72-hour incident reporting (vs. California’s 15 days); 24-hour notice for imminent risk.
- First-instance penalties cut from $10M to $1M.
- Adds quarterly catastrophic-risk summaries, whistleblower protection, and academic/Empire-AI carve-outs.
Executive order
Calif.
Effective Jul 28 ’26
Mar 30, 2026
California orders AI safety certifications for state procurement
Directs agencies to recommend certifications vendors can earn to prove their AI systems are safe to buy.
Exec. Order N-5-26▸ details
◆ Effective Jul 28, 2026
- Certification focus: illegal-content prevention (incl. CSAM), bias governance, and civil-rights safeguards.
- Also seeks contractor-responsibility reforms.
- Agency recommendations due July 28, 2026.
Signed into law
Ore.
Effective Jan 1 ’27
Mar 31, 2026
Oregon enacts an AI-companion law
Regulates emotion-simulating AI companions — banning reward loops and simulated distress, with a private right of action.
S.B. 1546▸ details
◆ Effective Jan 1, 2027
- Targets bots that retain memory and sustain human-like personal relationships.
- Prohibits engagement-maximizing reward systems and simulated distress.
- $1,000 per violation or actual damages; effective Jan 1, 2027.
Apr2026
5 items
Signed into law
Idaho
Effective Jul 1 ’27
Apr 1, 2026
Idaho enacts the Conversational AI Safety Act
Public-facing conversational AI must run self-harm detection with crisis referrals, curb sexual content, and post persistent disclaimers.
S.B. 1297▸ details
◆ Effective Jul 1, 2027
- Explicit developer safe harbor for third-party operator violations — a point of divergence from Iowa.
- Persistent disclaimer at session start and at least every three hours.
- AG-only enforcement: $1,000 per violation up to $500,000; effective Jul 1, 2027.
Federal report
U.S.
Apr 15, 2026
White House economic report leans into “AI dominance”
A CEA chapter tallies projected GDP gains from existing policy and pushes deregulation and data-center buildout.
Economic Report of the President▸ details
- Estimates gains from the National Policy Framework, OBBBA, and 2025 executive orders.
- No major new proposals — a case for staying the deregulatory course.
Signed into law
Neb.
Apr 17, 2026
Nebraska enacts an AI Safety Act
Joins the 2026 chatbot wave: companion AIs must post a persistent disclaimer or notify users at session start and every three hours.
L.B. 525▸ details
- Tracks the Idaho / Oregon / Washington template.
- AG-enforced, with no private right of action.
Federal guidance
U.S.
Apr 17, 2026
Banking regulators refresh model-risk guidance
The three federal banking agencies update model-risk-management principles — expressly excluding generative and agentic AI.
OCC / FRB / FDIC▸ details
- Scope is limited to non-generative, non-agentic models.
- Leaves GenAI and agentic systems for separate treatment.
Litigation
Colo.
Apr 24, 2026
Colorado’s 2024 AI Act stalls in court
After X.AI sued, the state agreed not to enforce the 2024 law — or any successor — until rulemaking concludes.
X.AI v. Weiser (D. Colo.)▸ details
- The 2024 S.B. 205 was scheduled to take effect June 30, 2026 after an earlier delay.
- X.AI sought an injunction on April 9; a joint enforcement pause followed April 24.
- Ultimately replaced by the narrower S.B. 189 (May 14).
May2026
5 items
Signed into law
Iowa
Effective Jul 1 ’27
May 2, 2026
Iowa enacts a conversational-AI law
Closely mirrors Idaho’s statute but drops the third-party safe harbor and broadens mental-health coverage.
S.F. 2417▸ details
◆ Effective Jul 1, 2027
- Covers both suicidal ideation and self-harm (Idaho covers ideation only).
- Uses a reasonable-person standard for mental-health claims and permits rulemaking.
- Effective Jul 1, 2027.
Takes effect
Utah
In force
May 6, 2026
Utah identity-protection law takes effect
Creates a licensable, descendible right to one’s own personal identity and a cause of action for its abuse.
S.B. 256▸ details
◆ Effective May 6, 2026
- The identity right runs 70 years past death and applies retroactively.
- AI authorship is no defense to a defamation claim; adds a notice step for such suits.
Signed into law
Colo.
Effective Jan 1 ’27
May 14, 2026
Colorado replaces its 2024 AI Act with a narrower ADMT law
Repeals S.B. 205’s duty-of-care and impact-assessment regime; targets automated decision-making behind consequential decisions.
S.B. 189▸ details
◆ Effective Jan 1, 2027
- “ADMT” is narrower than the old “high-risk AI” — it must process personal data.
- Consumers get notice, plus access, correction, and human-review rights.
- Drops the duty of reasonable care, mandatory risk programs, and annual impact assessments; effective Jan 1, 2027.
Executive order
Calif.
May 21, 2026
California braces its workforce for AI disruption
Directs state agencies to prepare for AI-driven labor-market and economic dislocation.
Exec. Order N-6-26▸ details
- Aimed at economic shocks from widespread AI adoption.
- Pairs with the state’s AI-in-the-workplace policy work.
Signed into law
Conn.
Effective Oct 1 ’26
May 27, 2026
Connecticut enacts a sweeping online-safety & AI law
An omnibus statute stitching together many of the year’s themes — chatbots, whistleblowers, provenance, employment, and more.
S.B. 5▸ details
◆ Effective Oct 1, 2026
- Companion-chatbot rules, a frontier whistleblower shield, and a regulatory sandbox.
- Employment-ADM limits, large-GenAI provenance duties, and subscription-service restrictions.
- A Surgeon-General-style social-media warning; core provisions effective Oct 1, 2026.
Jun2026
4 items
Signed into law
Conn.
Effective Oct 1 ’26
Jun 2, 2026
Connecticut adds GenAI subscription-notice rules
Large GenAI subscription services must get written acceptance before charging or renewing — and disclose usage caps.
H.B. 5222▸ details
◆ Effective Oct 1, 2026
- Applies to GenAI subscription services with 1M+ monthly users.
- Must disclose token/image/transcription limits and any right to throttle quality.
- A breach is an unfair-trade-practice violation; effective Oct 1, 2026.
Executive order
U.S.
Jun 2, 2026
Trump order sets up a “covered frontier model” review
Creates a capability-based frontier threshold and a framework for federal pre-release access to top models.
EO: Prompting Advanced AI Innovation & Security▸ details
- Federal government may access covered frontier models up to 30 days before broad release.
- Prioritizes AI-related crime and cyber defense.
- Expressly does not authorize mandatory licensing or preclearance.
Takes effect
Miss.
In force
Jun 8, 2026
Mississippi says architects can’t rubber-stamp AI
Updated board rules require an architect’s full knowledge of and involvement in preparing documents.
Miss. Bd. of Architecture▸ details
◆ Effective Jun 8, 2026
- Reviewing and approving AI-generated output is explicitly declared insufficient.
- Signals how licensed professions may police AI-assisted work.
Takes effect
N.Y.
In force
Jun 9, 2026
New York synthetic-performer disclosure takes effect
Advertisements must disclose AI-generated synthetic performers, subject to exceptions.
Synthetic Performer Law (2025)▸ details
◆ Effective Jun 9, 2026
- Applies to those who produce or create the advertisement.
- Part of New York’s broader synthetic-media disclosure push.
Jul2026
2 items
Proposed
U.S.
Jul 1, 2026
FTC floats an “accuracy suppression” deception theory
Proposes that trimming AI accuracy to serve ideological or “equity” aims could be deceptive under FTC Act §5.
FTC proposed policy statement▸ details
- Frames compliance-driven output-shaping — e.g. for state antidiscrimination laws — as a potential §5 violation.
- A safe harbor: clearly and repeatedly disclose that outputs prioritize objectives over user requests.
- Public comment open through July 31, 2026.
Reports due
Calif.
Effective Jul 28 ’26
Jul 28, 2026
California AI-procurement recommendations come due
Agency reports on safety certifications and contractor-responsibility reforms are due.
per Exec. Order N-5-26▸ details
◆ Effective Jul 28, 2026
- Due 120 days after Executive Order N-5-26.
- Feeds the state’s eventual procurement-certification scheme.
Aug2026
1 item
Takes effect
Calif.
Effective Aug 2 ’26
Aug 2, 2026
California AI Transparency Act takes effect
Large AI systems must provide user disclosures and offer AI-detection tools.
CA AI Transparency Act▸ details
◆ Effective Aug 2, 2026
- A 2025 amendment moved the date from January to August 2, 2026.
- Focuses on helping the public identify AI-generated content.
No measures match those filters.
Enacted lawExecutive orderFederal actionProposed / draftGuidanceEffective-date marker
Questions this page answers
＋Which US states have AI laws taking effect in 2026?
On January 1, 2026, four measures took effect at once: Illinois' civil-rights AI amendments (H.B. 3773), the Texas Responsible AI Governance Act (TRAIGA), California's frontier training-data disclosures (TFAIA), and California's chatbot guardrails (S.B. 243). They were followed by New York's RAISE Act (March 19 — reworked days later by S.8828), Utah's identity-protection law S.B. 256 (May 6), Mississippi's architecture-board AI rules (June 8), New York's synthetic-performer disclosure law (June 9), the California AI Transparency Act (August 2), and Connecticut's omnibus S.B. 5 and H.B. 5222 (October 1).
＋Is there a comprehensive federal AI law in 2026?
No. Federal activity so far consists of executive action (EO 14365's preemption push and the June 2026 covered-frontier-model order), agency guidance (Treasury's Financial Services AI RMF and Lexicon; the OCC/FRB/FDIC model-risk refresh), one proposed FTC policy statement, and pending bills — the TRUMP AMERICA AI Act discussion draft and the DEFIANCE Act. The operative compliance obligations in 2026 are overwhelmingly state law.
＋What are AI companion-chatbot laws, and which states have them?
They regulate AI systems that simulate human conversation and relationships: clear non-human disclosure (typically at session start and every three hours), self-harm and suicide-risk detection with crisis referrals, and stricter rules for minors. The 2026 roster: California (S.B. 243, in force), Washington (H.B. 2225, eff. Feb 1, 2027), Oregon (S.B. 1546, eff. Jan 1, 2027), Idaho (S.B. 1297, eff. Jul 1, 2027), Nebraska (L.B. 525), Iowa (S.F. 2417, eff. Jul 1, 2027), and Connecticut (S.B. 5, eff. Oct 1, 2026). The divergences matter — Idaho grants developers a third-party safe harbor; Iowa deliberately drops it.
＋What happened to the Colorado AI Act?
Colorado's 2024 AI Act (S.B. 205) never took effect as written. After X.AI sued (X.AI v. Weiser, D. Colo.), the state agreed in April 2026 not to enforce the law pending rulemaking, and on May 14, 2026 the legislature replaced it with the narrower S.B. 189 — an automated-decision-making (ADMT) law granting consumers notice, access, correction, and human-review rights while dropping the duty of reasonable care, mandatory risk programs, and annual impact assessments. It takes effect January 1, 2027.
＋What are the next AI compliance deadlines?
From July 18, 2026: California agency reports under Executive Order N-5-26 come due July 28, 2026; the California AI Transparency Act takes effect August 2, 2026; Connecticut's S.B. 5 and H.B. 5222 follow on October 1, 2026; Colorado S.B. 189 and Oregon S.B. 1546 on January 1, 2027; Washington H.B. 1170 and H.B. 2225 on February 1, 2027; and Idaho S.B. 1297 and Iowa S.F. 2417 on July 1, 2027. The countdown section above tracks these live.
＋Who maintains this tracker — and can I rely on it?
The tracker is maintained by Fidem Praxis, a practice at the seam of corporate law, management consulting, and machine learning, founded by Wasim M. Amin, J.D., M.B.A. It is compiled from public legislative and regulatory records and updated as measures move. Summaries are original paraphrases for orientation only — not legal advice. Verify against the primary sources, or talk to us about what a measure means for your deployment.
Methodology & citation
Scope. This tracker follows United States AI-governance measures active or advancing from January 1, 2026: enacted state statutes, executive orders, federal agency actions and guidance, litigation that changes a law's operative status, and the federal backdrop shaping preemption. Each entry records the instrument, jurisdiction, action date, effective date, subject area, and the provisions that matter in practice.
Sources & cadence. Compiled from public legislative and regulatory records — session laws, executive orders, agency releases, and dockets. "In force" status and the countdowns are assessed live against the date you open this page; the dataset itself is revised as measures move.
Limits. Summaries are original paraphrases for orientation, not legal advice, and effective dates and pending measures change. Verify against the primary sources before relying on any provision. Attorney Advertising.
Cite this page: Fidem Praxis, US AI Regulation Tracker 2026 (updated July 18, 2026), https://fidempraxis.com/ai-regulation-tracker
Counsel · Strategy · AI
The rulebook keeps moving. Your deployment shouldn't wait for it.
Fidem Praxis designs, deploys, and sustains governed AI systems inside regulated work — built to clear the Counsel, Strategy, and Model lenses on day one, and re-cleared as the law moves.
Founded by Wasim M. Amin, J.D., M.B.A. — 22 years of cross-border practice; architect of the training-data architectures frontier legal models are built on.