US State AI Compliance:
2026-2027 Deadline Calendar
Track AI governance deadlines across US states. Covers Colorado, Illinois, California, New York, Texas, and more. Updated as laws evolve.
State-by-State Deadline Table
| State / Jurisdiction | Law | Deadline |
|---|---|---|
| New York City |
Local Law 144
Annual bias audits for AI hiring tools. Fines:
$375-$1,500/violation.
|
In Effect |
| Illinois |
HB 3773
AI employment decision disclosure. Amends Illinois Human
Rights Act.
|
Jan 1, 2026 |
| California |
SB 53 (TFAIA)
Frontier AI safety reports and catastrophic-risk plans for
large models.
|
Jan 1, 2026 |
| California |
AB 2013
Training data transparency for generative AI. Disclosure
requirements.
|
Jan 1, 2026 |
| Texas |
TRAIGA
Governance docs, transparency disclosures, and high-risk AI
controls.
|
Jan 1, 2026 |
| Indiana |
ICDPA
Consumer data privacy. Applies to AI systems processing
personal data.
|
Jan 1, 2026 |
| Kentucky |
KCDPA
Consumer data privacy. Affects AI systems using Kentucky
resident data.
|
Jan 1, 2026 |
| Rhode Island |
RIDTPPA
Data transparency and privacy. Applies to AI data
processing.
|
Jan 1, 2026 |
| California |
SB 942
AI-generated content watermarking and visible disclosure.
|
Aug 2, 2026 |
| Colorado |
ADMTA (SB 26-189)
Automated decision-making transparency. Replaces repealed SB
24-205.
|
Jan 1, 2027 |
What This Means for AI Teams
State AI laws do not respect company headquarters. If your AI system affects users, employees, or customers in a regulated state, that state's laws apply. The compliance burden stacks: a national AI product may need to satisfy Colorado's transparency requirements, Illinois's employment disclosures, California's training-data reporting, and NYC's bias audits simultaneously.
Most state AI laws share a common DNA: they require some combination of risk assessment, transparency disclosure, human oversight, and audit trails. HUMMBL's governance primitives — kill switches, circuit breakers, delegation tokens, and append-only audit logs — address these requirements across jurisdictions using the same stdlib-only Python code.
Map These Deadlines to Your System
HUMMBL provides open-source governance primitives designed for multi-jurisdiction AI compliance. Kill switches for emergency halt, circuit breakers for failure isolation, delegation tokens for runtime attribution, and append-only audit logs for evidence preservation. All Python stdlib-only.
Frequently Asked Questions
What is the Colorado AI Act deadline?
Colorado's Automated Decision-Making Technology Act (SB 26-189) takes effect January 1, 2027. It replaces the repealed SB 24-205 and establishes a notice-based transparency framework requiring technical documentation and consumer disclosures for automated decision-making systems.
Which states have AI laws already in effect?
As of early 2026, Illinois HB 3773 (employment AI discrimination), California SB 53 (frontier AI safety), California AB 2013 (training data transparency), and NYC Local Law 144 (AI hiring bias audits) are all in effect. Texas TRAIGA is also in effect as of January 1, 2026.
Does my AI system need to comply with every state law?
State AI laws typically apply based on where your users or employees are located, not where your company is headquartered. If you have users, employees, or customers in a state with AI regulations, those laws may apply to your AI systems. The compliance burden stacks across jurisdictions.
What is the penalty for non-compliance with state AI laws?
Penalties vary by state and law. Colorado ADMTA penalties are enforced by the Attorney General under the Colorado Consumer Protection Act. Illinois HB 3773 creates private rights of action. NYC Local Law 144 fines range from $375-$1,500 per violation. California penalties vary by specific statute.
How can HUMMBL help with multi-state AI compliance?
HUMMBL provides open-source governance primitives that map to requirements across multiple state AI laws: kill switches for human oversight, circuit breakers for operational safety, delegation tokens for audit trails, and append-only audit logs for evidence preservation. All are Python stdlib-only.