New York Amends the RAISE Act to Align More Closely with California Law

New York has amended its sweeping frontier AI safety law, the Responsible AI Safety and Education Act (the “RAISE Act”), bringing it into closer alignment with California’s Transparency in Frontier Artificial Intelligence Act (CA SB-53) and potentially easing some multistate compliance challenges for large frontier developers.

The amendment changes the RAISE Act in three ways:

• The core transparency requirements of the RAISE Act remain largely intact, but the amendment narrows the scope of frontier developers classified as “large frontier developers” that are subject to additional obligations. In particular, the amendment replaces the prior compute-based definition of a “large frontier developer” with a revenue-based threshold identical to the one in CA SB-53 (i.e., developers that exceeded $500 million in annual gross revenue in the preceding calendar year).
• The amendment also reduces civil penalties that the New York Attorney General can seek from $10 million to $1 million for a first violation and from $30 million to $3 million for subsequent violations, bringing the penalty framework closer to CA SB-53’s penalties, which are capped at $1 million per violation.
• Enforcement of the law is now delayed until January 2027, giving developers more time to assess and comply with their obligations under the RAISE Act’s revised scope.

This alert summarizes the RAISE Act, including as amended by Senate Bill S8828; highlights notable differences between the RAISE Act and CA SB-53; and provides key takeaways for AI developers.

Scope and Applicability

The RAISE Act, which applies only to frontier models that are developed, deployed, or operated in whole or in part in New York State, defines a frontier model as a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations. The quantity of computing power includes the original training run and any subsequent modifications to a preceding foundation model.

Large frontier developers are defined as frontier developers that together with their affiliates collectively had annual gross revenues in excess of $500 million in the preceding calendar year.

Accredited colleges and universities engaged in academic research, as well as the Empire AI Consortium and Institute, are exempt from the RAISE Act.

Comparison to CA SB-53: Both laws use the same compute threshold to define a frontier model and the same revenue threshold to define a large frontier model developer. The RAISE Act, however, includes an explicit carve-out for academic institutions, which SB-53 does not. In addition, the RAISE Act limits its application to models developed, deployed, or operated, in whole or in part, in New York State, whereas SB-53 does not expressly restrict its scope to developers based in California and therefore may have extraterritorial application.

Key Requirements

As explained below, covered frontier developers, regardless of annual revenue, are subject to certain transparency and critical safety incident reporting requirements. However, large frontier developers are subject to more extensive requirements, such as risk assessments and government disclosures.

Transparency Requirements

Frontier AI Framework

A large frontier developer must write, implement, comply with, and conspicuously publish on its website a frontier AI framework that applies to its frontier models. Frameworks must be reviewed and, as appropriate, updated at least once per year, and material modifications must be published with a justification within 30 days. The frontier AI framework must include how the large frontier developer handles all of the following:

• Incorporating national standards, international standards, and industry consensus best practices into its framework;
• Defining and assessing thresholds used by the large frontier developer to identify and assess whether a frontier model has capabilities that could pose a catastrophic risk, including multitiered thresholds;
• Applying mitigations to address the potential for catastrophic risks based on the results of assessments undertaken pursuant to the RAISE Act;
• Reviewing assessments and adequacy of mitigations as part of the decision to deploy a frontier model or use it extensively internally;
• Using third parties to assess the potential for catastrophic risks and the effectiveness of mitigations;
• Revisiting and updating the frontier AI framework, including criteria that trigger updates and how the large frontier developer determines when its frontier models are substantially modified enough to require disclosures;
• Implementing cybersecurity practices to secure unreleased model weights from unauthorized modification or transfer by internal or external parties;
• Identifying and responding to critical safety incidents;
• Instituting internal governance practices to ensure implementation of these processes; and
• Assessing and managing catastrophic risk resulting from internal use of its frontier models, including risks resulting from a frontier model circumventing oversight mechanisms.

Transparency Report

Before, or concurrently with, deploying a new or substantially modified version of an existing frontier model, the RAISE Act requires:

• A frontier developer to clearly and conspicuously publish on its website a transparency report including the internet website of the frontier developer; a mechanism that enables a natural person to communicate with the frontier developer; the release date of the frontier model; and the intended uses of the frontier model. A published system card or model card that includes the necessary items satisfies the requirement. While redactions are permitted to protect trade secrets, cybersecurity, public security, or national security, the frontier developer must describe the character of, and justification for, the redactions in the report and retain the unredacted information for five years; and
• A large frontier developer to add summaries of specified information in its transparency report, including assessments of catastrophic risks from the frontier model conducted pursuant to the large frontier developer’s frontier AI framework; the results of those assessments; and the extent to which third-party evaluators were involved.

Comparison to CA SB-53: The two laws mandate identical frontier AI frameworks and transparency reports. The laws utilize the same definition of catastrophic risk.

Reporting Potential Critical Safety Incidents

The RAISE Act creates an oversight office (the “Office”) within the New York Department of Financial Services to oversee implementation and enforcement of the law, including the establishment of two reporting channels:

• Critical Safety Incidents. The Office must establish a mechanism for frontier developers and members of the public to report critical safety incidents. The reports must capture: (i) the date of the critical safety incident; (ii) the reasons the incident qualifies as a critical safety incident; (iii) a short and plain statement describing the incident; and (iv) whether the incident was associated with internal use of a frontier model.
• Confidential Risk Assessments. Every three months, large frontier developers are required to submit to the Office a written summary of any assessment of catastrophic risk resulting from internal use of its frontier models. The Office shall establish a mechanism to be used by large frontier developers to confidentially submit the required summary.

A frontier developer must report any critical safety incident pertaining to one or more of its frontier models to the Office within 72 hours of a determination that a critical safety incident has occurred or within 72 hours of the developer learning facts sufficient to establish a reasonable belief that a safety incident has occurred.

If a frontier developer discovers that a critical safety incident poses an imminent risk of danger of death or serious physical injury, it must disclose that incident within 24 hours to an authority, including any law enforcement agency or public safety agency with jurisdiction, that is appropriate based on the nature of the incident and as required by law. Reports are exempt from public disclosure, and the Office may share reports of critical safety incidents with other governmental entities at its discretion.

Beginning January 1, 2028, the Office must produce an annual report that includes anonymized and aggregated information about critical safety incidents that have been reviewed by the Office, but not information that would compromise the trade secrets of cybersecurity of the frontier developer. The Office must submit the report to specified authorities, including the governor. The Office may designate other federal laws, regulations, or guidance documents that impose or state standards or requirements for critical safety incident reporting that are “substantially equivalent” to or stricter than those required by the RAISE Act or that are intended to assess, detect, or mitigate the catastrophic risk. A frontier developer that intends to comply with the critical safety incident reporting requirements of the RAISE Act by complying with the requirements of such designated documents must declare its intent to do so to the Office. After doing so, the frontier developer will be deemed in compliance with the critical safety incident reporting requirements of the RAISE Act if it complies with the requirements. Failure to comply with the requirements will constitute a violation of the critical safety incident reporting section of the RAISE Act.

Comparison to CA SB-53: The RAISE Act imposes a shorter reporting window of 72 hours following a critical safety incident in comparison with SB-53’s 15-day window. SB-53 additionally includes whistleblower protections for employees of covered entities, which the RAISE Act does not.

Large Frontier Developer Disclosure Statements

Large frontier developers may not develop, deploy, or operate a frontier model in New York without filing a current disclosure statement with the Office and paying a pro rata assessment to fund program administration. Disclosure statements must be filed every two years and must include:

• The identity of the large frontier developer and all names under which it conducts business; the address of the principal place of business and the address of each office it maintains in New York State; if the developer or its ultimate parent is a privately or closely held company, a list of all entities that beneficially own a 5% or greater interest in the developer at the time of filing the disclosure statement, and a list of persons who formerly beneficially owned a 5% or greater interest in such owner or its predecessors during the preceding five years. If the owner or ultimate parent is a publicly traded company, a list of all persons or entities that beneficially own a 50% or greater interest in the large frontier developer at the time of registration; and the name and contact information of three points of contact.

The Office is required to publish a public list of large frontier developers that have filed disclosures, excluding contact information.

Comparison to CA SB-53: The RAISE Act adds a frontier developer disclosure obligation that requires large frontier developers to maintain current filings with the Office. SB-53 does not include this requirement.

Enforcement and Penalties

The Attorney General may seek civil penalties of up to $1 million for a first violation and $3 million for subsequent violations for failures relating to publication, reporting, false statements, or noncompliance with a developer’s own frontier AI framework. The statute expressly disclaims any private right of action.

If any person develops, deploys, or operates a large frontier model in part in New York State without a current disclosure filed with the Office pursuant to the RAISE Act, the Office may, after notice and hearing, impose a civil penalty of $1,000 for each day the entity fails to file a disclosure and an amount equal to the assessments owed.

The Office is granted broad rulemaking authority to implement the statute and may impose additional reporting or publication requirements to further safety and transparency objectives.

Comparison to CA SB-53: Both laws empower their state attorneys general to pursue civil enforcement, but the RAISE Act authorizes higher penalties of $1 million for a first violation and $3 million for subsequent violations for failures relating to publication, while SB-53 caps penalties at $1 million per violation. The RAISE Act establishes an office within the New York Department of Financial Services to oversee enforcement, whereas SB-53 requires that incident reports be submitted to the California Office of Emergency Services.

Key Takeaways

Companies that do business in New York and develop foundation models should consider the following steps to prepare ahead of time for the RAISE Act entering into force:

• Assess whether they qualify as a frontier developer or large frontier developer under the RAISE Act;
• Prepare for compliance with the RAISE Act’s transparency obligations, including developing a frontier AI framework subject to at least annual review;
• Evaluate existing AI governance, risk management, and documentation practices to determine whether current frameworks can be leveraged to meet the RAISE Act’s requirements; and
• Implement processes to identify events that may constitute critical safety incidents and establish procedures for timely reporting in accordance with the RAISE Act.

Maya Vishwanath, an AI Analyst at Morrison Foerster, contributed to this alert.