Automated decision systems.

Senator Padilla's comprehensive artificial intelligence legislation would establish new oversight requirements for high-risk automated decision systems used in critical domains like employment, housing, and healthcare. The bill requires developers and deployers of these systems to conduct impact assessments examining potential discrimination risks before implementation and mandates transparency around AI-driven decisions affecting individuals.

Under the proposal, developers must evaluate their systems' intended purposes, data inputs, and potential discriminatory impacts before deployment. Organizations using high-risk AI systems must notify individuals when such systems inform decisions about them, explaining the system's purpose, data used, and providing contact information. The bill also requires deployers to maintain public documentation about their AI systems and offer human review options for automated decisions.

The legislation creates enforcement mechanisms through the Attorney General and Civil Rights Department, who can pursue civil penalties ranging from $2,500 to $25,000 per violation. State agencies would be prohibited from contracting with entities that have violated the bill's provisions or other specified civil rights laws. Companies with 50 or fewer employees and AI systems already certified under equivalent federal standards would be exempt from the requirements.

The bill requires developers and deployers to establish governance programs with administrative and technical safeguards against algorithmic discrimination, scaled to factors like organization size and system risks. Impact assessments provided to state agencies or enforcement bodies must remain confidential to protect trade secrets while ensuring oversight of high-risk AI applications.