Senators Coons, Britt, Schatz, and Lankford introduce bipartisan AI safety and transparency legislation

Senators Chris Coons (D-Del.), Katie Britt (R-Ala.), Brian Schatz (D-Hawaii) and James Lankford (R-Okla.) introduced the AI Systems Transparency Act (ASTA) which would require AI companies to be more transparent with the public about what data their models are collecting from users, the safeguards they’ve put in place for children and adults, and the guardrails they’ve created to stop systems from going rogue or being used dangerously by bad actors. Under ASTA, the Federal Trade Commission (FTC) would have the power to enforce these transparency requirements, while ensuring that companies can continue to grow and innovate. ASTA would apply to closed and open-source models. Additionally, It would:

• Establish criteria for the size and scope of companies over which the FTC would gain jurisdiction
• Require disclosure of information about structure and policies typically included on “model cards”
• Require disclosure of preventative safeguards used for child safety, mental health, privacy, cybersecurity, disaster risk, and autonomous loss-of-control 
• Require disclosure of information relating to common violations of policies
• Require companies to update all disclosures with each new model or substantial updates to an existing model 
• Require both a consumer-facing disclosure and a disclosure that can be of use to independent researchers and evaluators