1. United States
2. Del.
3. Letter

1. Mandatory third-party safety evaluation before deployment of frontier AI models, particularly for cyber, biological, and autonomous-capability risks. 2. Incident reporting requirements so containment failures and safety breaches are disclosed to regulators, not just discovered after the fact. 3. An independent federal body with technical expertise and authority to set and enforce binding safety standards, similar in spirit to FDA or FAA oversight of other high-risk technologies. 4. Liability clarity so companies bear responsibility for harms caused by systems they release. AI companies are already sounding alarms about their own products. Legislators shouldn't wait for a worse incident before acting. I would welcome the opportunity to discuss this further and thank you for your attention to this urgent matter.