Frontier AI Framework
Documented policies by developers detailing how they manage and mitigate catastrophic AI risks.
How the laws define this
Both instruments agree that a frontier AI framework requires large frontier developers to establish and publish a documented set of measures detailing how they manage and mitigate catastrophic risks associated with their models. The primary divergence lies in ongoing compliance duties and required baseline topics. Under An Act to amend the general business law, in relation to transparency and safety requirements for developers of artificial intelligence frontier models, the framework consists generally of documented policies, procedures, and technical measures. Conversely, the Artificial Intelligence Safety Measures Act expands these obligations by requiring developers to implement and annually update the framework. It also explicitly mandates specific components, outlining methodologies for catastrophic-risk assessment, mitigation strategies, cybersecurity, internal governance, third-party evaluations, and the management of risks stemming from the internal use of frontier models.
Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| New York, United States | An Act to amend the general business law, in relation to transparency and safety requirements for developers of artificial intelligence frontier models | In Force | This definition emphasizes transparency and safety requirements, mandating publication of policies, procedures, and technical measures, while focusing on risk management and incident reporting. |
| Illinois, United States | Artificial Intelligence Safety Measures Act | Adopted | This definition expands to include annual updates, detailed methodologies for risk assessment, cybersecurity, governance, third-party evaluations, and risks from internal model use. |
What it triggers
- In New York, United States, large frontier developers are required to establish and publish documented policies, procedures, and technical measures detailing how they manage and mitigate catastrophic risks.
- Under the Illinois Artificial Intelligence Safety Measures Act, large frontier developers are required to establish, implement, publish, and annually update a detailed framework outlining risk assessment, mitigation, cybersecurity, governance, third-party evaluations, and internal risks.source
Definitions (2)
Documented policies, procedures, and technical measures that large frontier developers are required to establish and publish. This framework details how they manage and mitigate catastrophic risks associated with their AI models.
A detailed framework that large frontier developers are required to establish, implement, publish, and annually update. It outlines methodologies for catastrophic-risk assessment, mitigation strategies, cybersecurity, internal governance, third-party evaluations, and managing risks from internal use of frontier models.
Questions and answers
What must be included in a Frontier AI Framework under New York law?
It includes documented policies, procedures, and technical measures for managing and mitigating catastrophic risks. The focus is on transparency, safety requirements, risk management, and incident reporting.
What are the ongoing obligations for Frontier AI Frameworks in Illinois?
Large frontier developers must establish, implement, publish, and annually update the framework. It covers catastrophic-risk assessment, mitigation strategies, cybersecurity, internal governance, third-party evaluations, and internal use risks.
How do Frontier AI Frameworks differ between New York and Illinois?
New York's version mandates policies, procedures, and technical measures with a focus on transparency and incident reporting. Illinois requires annual updates and specifics like third-party audits, cybersecurity, and internal governance.