EveryAILaw.com

Requirements at a glance

This regulation imposes 6 specific requirements for Transparency & Disclosure across 1 provision:

• Publish frontier AI framework — LFDs must write, implement, and clearly publish a frontier AI framework covering governance structures, catastrophic risk mitigation, cybersecurity practices, and standards alignment (§ 22757.12)
• Annual review — Review the framework at least annually and update it as appropriate (§ 22757.12(b)(1))
• Material modification — Publish a materially modified framework and a justification for the modification within 30 days (§ 22757.12(b)(2))
• Model transparency reports — Before or when deploying a new or substantially modified frontier model, publish the model information required by § 22757.12(c); large frontier developers must also publish summaries of catastrophic-risk assessments, results, third-party evaluator involvement, and framework implementation steps
• Internal-use risk summaries — Large frontier developers must transmit summaries of catastrophic-risk assessments from internal model use to OES every three months, or on another reasonable schedule communicated in writing (§ 22757.12(d))
• Redaction allowance — Frontier developers may make necessary redactions to protect trade secrets, cybersecurity, public safety, or national security, or comply with law; describe their character and justification as permitted and retain unredacted information for five years (§ 22757.12(f))