Policy
California ADMT Rules Take Effect January 1, 2027
Businesses using automated decision-making tools for employment, housing, and other significant choices face new notice and opt-out requirements under CCPA.
Key takeaways
- California CCPA regulations for automated decision-making technology take effect January 1, 2027, covering employment, finance, healthcare, housing, and education decisions.
- Businesses must provide pre-use notices explaining ADMT purpose, data processing, opt-out procedures, and human oversight roles.
- Companies can avoid opt-out requirements only by offering human appeal processes or proving their tools don't discriminate based on protected characteristics.
- Violations carry penalties up to $2,633 per violation or $7,988 per intentional violation, enforced by California's Attorney General and Privacy Protection Agency.
- Organizations already using ADMT have until January 1, 2027 to implement compliance workflows including risk assessments.
New compliance deadline for automated decision tools
Businesses subject to the California Consumer Privacy Act face a January 1, 2027 deadline to comply with new regulations governing automated decision-making technology. The rules apply when companies use ADMT to make significant decisions affecting employment, finances, healthcare services, housing, or education.
The regulations impose specific obligations on employers using AI-powered tools for recruiting, resume screening, applicant scoring, video interview analysis, performance ratings, scheduling, promotions, disciplinary actions, and terminations. Job applicants, employees, and independent contractors must receive notice before these technologies are deployed.
Core requirements under the new framework
The ADMT regulations establish three primary obligations: pre-use notice, opt-out rights, and access rights.
The pre-use notice must describe in specific terms how the technology will be used, explain its purpose, identify the types of personal information processed, clarify how consumers can opt out (or whether an exception applies), and detail any human involvement in decision-making. These notices must be accessible to people with disabilities.
Businesses must generally allow consumers to opt out of automated decision-making, though the law carves out narrow exceptions. Companies can bypass the opt-out requirement if they provide an appeal process to a human reviewer with authority to overturn the tool's decision. Additional exceptions apply when ADMT is used solely to assess work performance or educational capability for admission or hiring purposes, or when allocating work and compensation—provided the technology operates without unlawful discrimination based on protected characteristics.
When an ADMT makes a significant decision, businesses must provide consumers with a clear explanation of the specific purpose, how personal information was processed to generate the output, and the decision's outcome.
Why it matters
These regulations represent California's most comprehensive attempt to regulate AI in high-stakes decision-making. Companies already using these tools have less than two months to implement compliance workflows, while the risk assessment requirement under CCPA adds another layer of operational complexity. The enforcement mechanism—with penalties reaching $7,988 per intentional violation—creates significant financial exposure for businesses that fail to prepare.
Enforcement and penalties
The California Attorney General and California Privacy Protection Agency will enforce these rules. For 2025-2026, violations carry fines up to $2,633 per violation or $7,988 per intentional violation. The CPPA adjusts penalty amounts biannually each January in odd-numbered years.
Businesses using ADMT before January 1, 2027 must achieve compliance by that date. Those deploying these technologies on or after January 1, 2027 must comply immediately.
Preparing for compliance
Companies should establish processes to handle opt-out requests, appeals to human reviewers, access requests, and discrimination complaints related to automated decision-making technology. Organizations considering new AI-powered software purchases should verify their ability to meet these regulatory requirements before acquisition.
These details were first reported by Ogletree Deakins in a blog post authored by Anne Yarovoy Khan, Sean P. Nalty, and Leah J. Shepherd.