Automated decision systems.

Assembly Member Bauer-Kahan's automated decision systems legislation establishes comprehensive oversight of artificial intelligence and machine learning technologies used to make consequential decisions affecting Californians. The bill creates a new regulatory framework requiring developers and deployers of these systems to conduct impact assessments, provide detailed disclosures, and enable appeals of automated decisions.

Under the measure, developers must evaluate covered automated decision systems before deployment and annually thereafter, documenting their purpose, approved uses, expected accuracy, and potential disparate impacts. Beginning in 2030, independent third-party auditors will assess compliance with these requirements. Developers must provide deployers with assessment summaries and technical documentation while maintaining records for five years after a system's use ends.

The legislation requires deployers to notify individuals before and after consequential decisions involving automated systems, explaining how the technology was used and what characteristics were evaluated. Subjects of these decisions have 30 days to correct inaccurate personal information or appeal outcomes. Deployers using systems that affect more than 6,000 people over three years without proper documentation assume developer responsibilities under the law.

The Attorney General and other specified public entities may bring civil actions for violations, with penalties up to $25,000 per incident. While the requirements apply broadly across sectors like employment, education, housing and healthcare, the bill exempts systems used solely for cybersecurity, national defense, fraud detection and certain other limited purposes. The measure also amends existing civil rights laws to make compliance with these new requirements relevant evidence in discrimination cases.