California is considering a new law (Assembly Bill 331), also known as the Automated Decision Systems Accountability Act.  Modeled after the Biden Administration’s Blueprint for an AI Bill of Rights (whitehouse.gov), AB 331 would control the use of machine-based systems in making “consequential” employment decisions such as compensation, promotions, hiring, termination, and automated task allocations.

If passed and signed into law, AB 331 would impact employers who use AI, even through third-party vendors, in three major ways:

1. Mandating Disclosure Before AI is Used

AB 331 would require companies to disclose not only their use of automated decision tools, but also explain their purpose and how AI is being used.  Where a consequential decision is being made solely on an automated decision tool, individuals may have the ability to opt out and request an alternative selection process or accommodation.

2. Requiring Annual Impact Assessments

Under the proposed legislation, employers would have to conduct annual impact assessments to identify and mitigate potential biases in their AI systems.  The impact assessment would have to include a statement of purpose for the AI and its intended benefits, a summary of the data collected, the extent the tool is consistent with from the developer’s statement of use, an analysis of the potential adverse impact on protected categories (e.g., sex, race, or English proficiency), a description of the safeguards implemented to mitigate risk, and how the AI will be evaluated for validity or relevance.

The outcome of these impact assessments would have to be reported to the California Civil Rights Department (formerly known as the Department of Fair Employment and Housing).  Failure to do so could result in a $10,000 fine for every day AI is used and a company has not submitted the assessment.

3. Allow Suits Against Employers for discriminatory impact of AI tools

The most recent proposed revisions to AB 331 also provide for a private right of action such that employees and applicants suffering a discriminatory impact on the basis of a protected category could file a lawsuit against companies using AI in their employment decisions.  Unsurprisingly, the bill would establish yet another one-sided attorneys’ fees award for prevailing plaintiffs, incentivizing individuals and their attorneys to bring lawsuits against California-based employers.

While the bill is still in the early stages of the legislative process and could face significant changes along the way (or even be scrapped altogether), as AI becomes more integrated into our everyday lives, employers should be prepared for laws like AB 331 to appear on the horizon with significant and complex restraints on employers in the coming years.  We’ll provide updates as they unfold.