Comment by Mary Edly-Allen

Requires large frontier developers to create, implement, publish, and annually update a frontier AI framework addressing catastrophic-risk assessment, mitigations, cybersecurity, internal governance, third-party evaluations, and risks from internal use of frontier models. Requires transparency reports before deploying new or substantially modified frontier models and requires summaries of catastrophic-risk assessments. Mandates annual independent third-party audits and establishes access, reporting, retention, and publication requirements for audit results.