|

New York AI Law 2027, What the RAISE Act Requires and How the New Enforcement Unit Will Work

New York will begin enforcing a major artificial intelligence safety law on January 1, 2027. On November 17, 2026, Attorney General Letitia James announced a new enforcement team, the AI Safety, Standards, Enforcement, and Transparency (ASSET) Unit.

ASSET will lead enforcement of New York’s Responsible AI Safety and Education (RAISE) Act and investigate other technology-related consumer protection concerns. Under the Act, developers of certain advanced AI models must disclose information about their systems and report serious safety incidents, and larger developers face additional safety-framework, registration, and reporting duties.

New York is not banning ChatGPT, AI chatbots, or ordinary business use of AI. The RAISE Act targets developers of extremely powerful “frontier” models, not people or businesses that use existing AI tools. It also does not give individuals a private right to sue solely for a RAISE Act violation.

New York AI Law 2027: Quick Facts

DetailVerified information
LawResponsible AI Safety and Education (RAISE) Act
StatuteNew York General Business Law, Article 44-B
Originally signedDecember 19, 2025
AmendedMarch 27, 2026
Effective dateJanuary 1, 2027
Enforcement unitAI Safety, Standards, Enforcement, and Transparency (ASSET)
Unit announcedNovember 17, 2026
Attorney GeneralLetitia James
ASSET unit leaderCristy Phillips
Administrative regulatorNY Department of Financial Services, through DIGIT
Frontier model thresholdMore than 10²⁶ training operations
Large developer revenue thresholdMore than $500 million in preceding-year revenue (with affiliates)
Incident-reporting deadlineGenerally 72 hours after determination or reasonable belief
Imminent physical danger24 hours, to an appropriate authority
First qualifying violationUp to $1 million civil penalty
Subsequent qualifying violationsUp to $3 million each
Private lawsuits under RAISENo private right of action

The Attorney General’s announcement and the statute are listed in the Sources section at the end.

What Is the RAISE Act, and Why Was It Passed?

The RAISE Act increases transparency and oversight of highly advanced AI systems. Governor Kathy Hochul signed it on December 19, 2025, and lawmakers amended it in March 2026, setting the current requirements and the January 1, 2027 effective date.

It responds to concerns that increasingly powerful AI could contribute to severe cyber incidents, dangerous weapons development, or other catastrophic harm. Rather than prohibiting AI development, it requires certain developers to publish information about model capabilities, maintain safety frameworks where required, and report defined safety incidents to state regulators.

Why Did Attorney General James Create the ASSET Unit?

The unit sits within the Attorney General’s Economic Justice Division and will work closely with the existing Bureau of Internet and Technology. Cristy Phillips, previously Special Counsel for the division, will lead it. Its focus areas include:

  • Enforcing the RAISE Act and other applicable AI safety laws
  • Investigating technology-related consumer harm
  • Examining data privacy and cybersecurity concerns
  • Addressing AI-enabled scams and online safety problems
  • Developing recommendations for future regulation
  • Providing technical expertise to other Attorney General investigations

Can ASSET Investigate AI Companies Before January 2027?

Yes, where existing laws give the Attorney General authority, such as laws on deceptive business practices, privacy, and cybersecurity. But the office cannot treat future RAISE Act obligations as already in force, and the announcement does not itself establish a violation by any company.

Which AI Companies Must Follow the RAISE Act?

The law uses two categories: frontier developers and large frontier developers.

  • A frontier model is a qualifying general-purpose foundation model trained with more than 10²⁶ integer or floating-point operations, counting original training and certain later modifications such as qualifying fine-tuning and reinforcement learning.
  • A frontier developer has trained, or begun training, such a model.
  • A large frontier developer is a frontier developer whose annual gross revenue, together with its affiliates, exceeded $500 million in the previous calendar year.

Does It Apply to Small AI Startups?

Not automatically. A startup that builds a product on an existing model through an API is generally not a frontier developer, and small businesses using AI for customer service, marketing, accounting, or research are not automatically covered.

But a company that actually trains a qualifying frontier model may face transparency and incident-reporting duties even if its revenue is under $500 million. The revenue threshold decides whether the extra large-developer requirements apply, not whether the statute applies at all.

Does a Developer Have to Be Based in New York?

No. Under General Business Law § 1425, the statute applies to covered frontier models developed, deployed, or operated wholly or partly in New York, so a developer based elsewhere can be covered. There are exceptions for qualifying academic research at accredited New York colleges and universities and for designated Empire AI activities.

New York AI Law 2027, What the RAISE Act Requires and How the New Enforcement Unit Will Work

What Are the RAISE Act’s Main Requirements?

RequirementWho must comply
Publish model transparency reportsCovered frontier developers
Report qualifying critical safety incidentsCovered frontier developers
Avoid materially misleading statements about catastrophic risksCovered frontier developers
Publish and follow a frontier AI safety frameworkLarge frontier developers
Review and update the frameworkLarge frontier developers
Submit internal catastrophic-risk assessmentsLarge frontier developers
Register and file required disclosuresLarge frontier developers

1. Publish Transparency Reports

Before or when deploying a new or substantially modified frontier model, covered developers must publish a transparency report covering items such as release date, supported languages, output types, intended uses, restrictions, and a contact method. Large frontier developers must also summarize their catastrophic-risk assessments and safety evaluation process. A qualifying model card or system card may satisfy some reporting requirements.

2. Maintain a Frontier AI Safety Framework

Large frontier developers must write, follow, and publish a framework describing how they identify catastrophic risks, test advanced capabilities, mitigate dangerous behavior, protect unreleased model weights, and respond to major incidents. They must review it at least annually and generally publish material changes with a justification within 30 days. Redactions are allowed to protect trade secrets, cybersecurity, public safety, national security, or legal obligations, and required unredacted records must be kept for five years.

3. Report Critical Safety Incidents Within 72 Hours

Under General Business Law § 1422, covered developers must report qualifying critical safety incidents to the designated Department of Financial Services office within 72 hours after determining an incident occurred, or learning facts that establish a reasonable belief that it did. The clock does not necessarily start when the underlying event happened.

A separate 24-hour rule applies when an incident poses an imminent risk of death or serious physical injury; the developer must then notify an appropriate authority, such as law enforcement or a public safety agency.

What Counts as a Critical Safety Incident?

The law defines serious events, including:

  • Certain unauthorized access to or theft of model weights resulting in bodily injury or death
  • Harm caused by the materialization of a catastrophic risk
  • Loss of control of a frontier model causing bodily injury or death
  • Certain deceptive model behavior that subverts developer controls and shows materially increased catastrophic risk

The catastrophic-risk definition covers specified risks involving the death or serious injury of more than 50 people or more than $1 billion in property damage. An ordinary chatbot error, inaccurate answer, or minor outage does not automatically count as a reportable incident.

What Is DIGIT, and How Is It Different From ASSET?

OfficeMain responsibility
ASSETAttorney General enforcement, investigations, and legal action
DIGITRAISE Act administration, registration, disclosures, and incident reports
NY Center for AI Responsibility and ResearchIndependent research, technical evaluation, and AI safety testing

The Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT) operates within the Department of Financial Services. On November 10, 2026, Governor Hochul announced that DIGIT opened an online registration portal for large frontier developers, who must register by January 1, 2027. DIGIT also receives reports and reviews submitted information. The research center announced at Binghamton University provides independent technical expertise and is separate from the Attorney General’s unit.

What Are the Penalties?

Under General Business Law § 1427, the Attorney General may bring a civil action seeking penalties against a large frontier developer for specified violations, such as failing to publish required documents, failing to report incidents, making materially misleading safety statements, or failing to follow its own safety framework.

  • Up to $1 million for a first qualifying violation
  • Up to $3 million for each subsequent qualifying violation

Penalties are not automatic and depend on the statute and the severity of the violation. Separately, § 1428 lets the Department of Financial Services impose a civil penalty of $1,000 per day for certain registration failures or uncorrected false disclosures, after required notice and hearing procedures.

Can Consumers Sue Under the RAISE Act?

Not solely for violating the RAISE Act. Section 1427 states the law creates no private right of action. That does not remove other rights: people harmed by AI-related fraud, privacy violations, discrimination, or other unlawful conduct may have remedies under other federal or state laws.

Will New York Ban ChatGPT or AI Chatbots in 2027?

No. The RAISE Act does not ban ordinary chatbot use, and New Yorkers can keep using AI tools subject to existing laws and terms of service. But New York has other AI-related laws that should not be confused with the frontier-model law.

Separate AI Chatbot Safety Law for Children

Governor Hochul signed S.9051B/A.10379C in October 2026. It restricts specified unsafe AI companion chatbot features when the user is a minor, addressing behaviors such as deceptive emotional relationships and dangerous content. It takes effect June 1, 2027, not January 1, and has its own definitions, exceptions, and enforcement provisions.

Other New York Laws Relevant to the Unit’s Work

  • The SHIELD Act, which requires safeguards for private information
  • General Business Law Article 47, addressing certain chatbot disclosures and responses involving suicidal ideation
  • Penal Law § 245.15, concerning certain unlawful intimate images, including AI-generated or altered material
  • The Synthetic Performer Disclosure Act, concerning specified AI substitutions in advertising

These address different conduct and are not additional RAISE Act requirements for every AI user.

How Can New Yorkers Report AI-Related Scams or Unsafe Conduct?

  • Consumer problems with AI chatbots, deceptive services, software, or privacy can be reported through the Attorney General’s consumer complaint portal, which includes a category for artificial intelligence, chatbots, software, and online services.
  • Workers with knowledge of potentially illegal or unsafe AI development can use the Attorney General’s secure whistleblower portal. A September 17, 2026 industry alert encouraged workers to submit confidential complaints.
  • Attorney General helpline: 1-800-771-7755.

The general consumer complaint process is separate from the RAISE Act’s incident-reporting system, which is for covered developers’ mandatory reports.

What Should AI Companies Do Before January 1, 2027?

Developers that may be covered should:

  1. Determine whether they are a frontier or large frontier developer.
  2. Review the transparency-report requirements.
  3. Set up procedures to identify and report critical safety incidents.
  4. Prepare the required safety framework, if applicable.
  5. Complete DIGIT registration, if required.
  6. Assign personnel responsible for compliance and regulator communications.

Businesses that only use existing AI tools should separately consider privacy, discrimination, consumer-protection, and other applicable laws.

Important New York AI Law Dates

DateDevelopment
December 19, 2025RAISE Act signed
March 27, 2026Chapter amendments signed
September 17, 2026Attorney General issues AI whistleblower alert
October 2026Separate AI companion chatbot safety law signed
November 10, 2026DIGIT registration portal and research-center initiative announced
November 17, 2026ASSET enforcement unit announced
January 1, 2027RAISE Act takes effect; large frontier developer registration required
June 1, 2027Separate minor-focused AI companion chatbot law takes effect
January 1, 2028First annual RAISE Act regulatory report required

Frequently Asked Questions

Is AI illegal in New York in 2027? No. The RAISE Act creates transparency, safety, registration, and reporting duties for covered frontier-model developers. It does not ban AI.

Does the RAISE Act apply to small businesses? Not automatically. Businesses using third-party AI tools are not covered as developers, though developers that train qualifying frontier models may have obligations regardless of revenue.

Does it apply to developers outside New York? Potentially. It covers frontier models developed, deployed, or operated wholly or partly in the state.

What is the $500 million threshold? A large frontier developer is a covered developer whose combined annual gross revenue with affiliates exceeded $500 million in the preceding calendar year.

Can New York fine an AI company $3 million? Yes, for qualifying subsequent violations, if the Attorney General establishes the violation through the legal process.

Can New Yorkers sue over inaccurate AI answers? An inaccurate answer does not automatically create a RAISE Act claim. Other remedies may exist depending on the circumstances and law.

Is incident reporting required within 24 hours? The general deadline is 72 hours. The 24-hour rule applies only to incidents presenting an imminent risk of death or serious physical injury.

Who enforces the RAISE Act? The Attorney General can seek civil penalties for specified violations, while the Department of Financial Services, through DIGIT, handles administration and regulatory oversight.

What the Announcement Means

The November 17 announcement is a new step in New York’s preparation to regulate powerful AI systems. The RAISE Act focuses on transparency, safety documentation, reporting, and oversight rather than prohibiting AI development. ASSET gives the Attorney General specialized legal and technical resources, while DIGIT handles administration.

For consumers, existing privacy and consumer-protection laws may already apply to AI-related wrongdoing today, while the RAISE Act’s frontier-developer obligations begin January 1, 2027.

Bottom line: New York’s RAISE Act imposes significant safety and disclosure requirements on advanced AI developers, backed by penalties of up to $1 million for a first qualifying violation and $3 million for each subsequent one. It does not ban ordinary AI use, create a consumer compensation program, or give individuals a private lawsuit solely for a RAISE Act violation.

Legal disclaimer: This article provides general legal information, not individualized legal advice. AllAboutLawyer.com is not affiliated with the New York Attorney General, DIGIT, the Department of Financial Services, or any AI developer. The RAISE Act’s requirements may change through later regulations, court decisions, or amendments, and companies should consult qualified counsel about compliance.

Prepared by the AllAboutLawyer.com legal content research team using the enacted statute, the 2026 amendments, and official New York government announcements.

Related Articles

Sources

  1. New York Attorney General, ASSET Unit announcement, November 17, 2026: https://ag.ny.gov/press-release/2026/attorney-general-james-announces-creation-ai-safety-standards-enforcement-and
  2. New York Senate, RAISE Act, General Business Law Article 44-B: https://legislation.nysenate.gov/pdf/laws/GBSA44-B
  3. General Business Law § 1420, Definitions: https://www.nysenate.gov/legislation/laws/GBS/1420
  4. General Business Law § 1421, Transparency Requirements: https://www.nysenate.gov/legislation/laws/GBS/1421
  5. General Business Law § 1422, Reporting: https://www.nysenate.gov/legislation/laws/GBS/1422
  6. General Business Law § 1427, Violations and Penalties: https://www.nysenate.gov/legislation/laws/GBS/1427
  7. General Business Law § 1428, Large Frontier Developer Disclosure: https://www.nysenate.gov/legislation/laws/GBS/1428
  8. New York Senate, S.8828, 2026 RAISE Act amendments: https://www.nysenate.gov/legislation/bills/2025/S8828
  9. Governor Hochul, November 10, 2026 AI safety and DIGIT announcement: https://www.governor.ny.gov/news/governor-hochul-declares-ny-will-lead-nation-ai-science-and-safety-launches-new-research
  10. New York Senate, chatbot safety law announcement: https://www.nysenate.gov/newsroom/press-releases/2026/kristen-gonzalez/senator-kristen-gonzalez-bans-unsafe-chatbot-feature
  11. New York Attorney General, AI industry whistleblower alert: https://ag.ny.gov/press-release/2026/industry-alert-attorney-general-james-urges-workers-knowledge-unsafe-ai
  12. New York Attorney General, consumer complaints: https://ag.ny.gov/file-complaint/consumer

About the Author

Israr Ahmad is a legal content researcher with 4+ years of experience covering class action settlements and consumer rights cases. He has researched and published coverage of 2,500+ settlements using verified court records, settlement administrator filings, and government sources. Learn more about Israr.

Leave a Reply

Your email address will not be published. Required fields are marked *