S.Amdt. 6247

This amendment, submitted by Ms. Hassan, aims to add a new section (SEC. 225) to bill S. 4784. Its purpose is to authorize the Secretary of Homeland Security to start prize competitions to improve two specific areas of artificial intelligence: interpretability and adversarial robustness.

Key Changes and Additions:

• Prize Competitions: The Secretary must start at least one competition for each of the following:

  • Interpretability: Focused on the science of how humans understand how AI models make decisions and respond to inputs.
  • Adversarial Robustness: Focused on developing AI models that can resist attacks meant to make them produce harmful, restricted, or incorrect outputs.

• Consultation and Administration: The Secretary is required to consult with various officials, including the Secretary of Commerce, the National Cyber Director, and industry experts. The Secretary may also partner with nonprofit, for-profit, or government agencies to run these competitions.

• Accountability and Reporting: Within 180 days after the first competition ends, the Secretary must report to the appropriate House and Senate committees on the results, identified gaps in the fields, and suggested actions for Congress.

• Funding: The amendment authorizes $10,000,000 to be appropriated for these activities from fiscal year 2027 through 2030.