New York Opens RAISE Act Registration: OpenAI, Anthropic Face Jan. 1 Deadline
New York opened RAISE Act registration on November 10, giving large AI developers until January 1, 2027 to register or face fines of up to $1 million.
Image: Wikimedia Commons / U.S. Secretary of Defense
ALBANY, New York — New York opened registration under its Responsible AI Safety and Education Act on Tuesday, November 10, giving the largest artificial intelligence developers until January 1, 2027 to sign up or face penalties of up to $1 million for a first violation, the Office of Governor Kathy Hochul said.
The RAISE Act, which Hochul signed in December 2025 and the state legislature amended in March, is among the most far-reaching AI safety laws in the United States. It requires large frontier developers to publish safety frameworks, report critical safety incidents within 72 hours and release public transparency reports for new models.
- Registration opened November 10, 2026. Large frontier developers must register by January 1, 2027, the day the law takes effect.
- It covers developers with more than $500 million in annual revenue training frontier models using more than 1026 operations.
- Covered firms must publish safety frameworks, report critical incidents within 72 hours and issue transparency reports. Fines reach $1 million for a first violation, $3 million for repeat ones.
- A new office, DIGIT, inside the Department of Financial Services will enforce the law. New York also launched an AI safety research centre at SUNY Binghamton.
Who Must Register Under the RAISE Act, and by When
The duty falls on large frontier developers: companies with annual revenue above $500 million, counting affiliates, that have trained or begun training a foundation model using more than 1026 operations. The test reaches only the biggest labs, including OpenAI and Anthropic.
Registration runs through the RAISE Portal, opened on November 10. The deadline is January 1, 2027, the day the law takes effect, leaving covered developers about seven weeks to get on record. The state has not named the companies it counts as covered.
What the RAISE Act Requires: Frameworks, 72-Hour Reports, Fines
From January, large developers must publish a safety framework on their websites, report a critical safety incident to the state within 72 hours and publish a transparency report for each new model. The reporting and transparency duties apply to all frontier developers.
Enforcement sits with DIGIT, the Office of Digital Innovation, Governance, Integrity, and Trust, a new office inside the Department of Financial Services. Violations carry civil penalties of up to $1 million for a first offence and $3 million for each later offence. Hochul said the portal gives the state a direct view of how the largest AI companies operate. “What this portal does is give us a line of sight, more visibility into the inner workings of these AI companies,” she said on November 10.
How New York’s AI Law Compares With Illinois
Illinois passed its own frontier AI safety law, SB 315, which also takes effect on January 1, 2027. The two laws match on the three core tests: the $500 million revenue threshold, the 72-hour incident-reporting window and penalty caps of $1 million for a first violation and $3 million for repeat ones. They part on oversight: Illinois requires independent third-party audits and whistleblower protections; New York’s law has neither. New York amended the RAISE Act in March 2026 to align it more closely with California’s frontier AI law.
| Requirement | New York | Illinois |
|---|---|---|
| Takes effect | January 1, 2027 | January 1, 2027 |
| Revenue threshold | $500 million+ | $500 million+ |
| Incident reporting | Within 72 hours | Within 72 hours |
| Maximum penalties | $1M first; $3M repeat | $1M first; $3M repeat |
| Third-party audits | Not required | Required |
| Whistleblower protections | Not included | Included |
Frequently Asked Questions
Do OpenAI and Anthropic have to register under New York’s RAISE Act?
Both fall in the group the law targets: large frontier developers with more than $500 million in annual revenue training models using more than 1026 operations. The state has not named the companies it counts as covered.
What happens if an AI company misses the January 1, 2027 RAISE Act deadline?
A company that fails to register, publish a safety framework or report a critical incident on time faces civil penalties of up to $1 million for a first violation and $3 million for each later one. The DIGIT office will enforce the law once it takes effect.
What counts as a frontier AI model under the RAISE Act?
A frontier model is a foundation model trained, or in training, using more than 1026 operations, including fine-tuning and reinforcement learning. Only developers above the $500 million revenue threshold must register.
Registration stays open through January 1, 2027, when the law takes effect and DIGIT’s enforcement powers begin.
