Automated decision systems.

Senator Padilla's artificial intelligence regulation bill establishes comprehensive oversight of high-risk automated decision systems in California, requiring impact assessments and governance programs to prevent algorithmic discrimination. The legislation mandates that developers and deployers evaluate AI systems used for critical decisions about education, employment, housing, healthcare and other essential services before implementation.

The bill creates a tiered compliance framework based on organization size and system risk level. Developers must conduct pre-deployment assessments examining potential discriminatory impacts and implement safeguards, while deployers must perform evaluations within two years of system adoption. State agencies can opt out of assessments when using unmodified systems as intended, provided developers meet contract requirements and no discrimination risks are identified.

When automated systems are used to make decisions about individuals, deployers must provide notice explaining the AI's purpose, data usage, and appeal rights. Organizations must maintain public disclosures about their high-risk AI deployments and risk management approaches. The Attorney General and Civil Rights Department can pursue civil penalties ranging from $2,500 to $25,000 per violation, with a 45-day cure period available based on factors like violation intent and entity resources.

The legislation exempts organizations with 50 or fewer employees and systems already approved under equivalent federal standards. It also prohibits state agencies from contracting for high-risk AI systems unless vendors certify compliance with California's civil rights and non-discrimination laws. Impact assessments submitted to oversight agencies remain confidential to protect trade secrets while ensuring accountability.