California Introduces Another AI Anti-Discrimination Bill: What Employers Need to Know
A newly proposed California bill seeks to regulate artificial intelligence (AI) decision-making tools in employment and other key areas. Introduced on February 20 by Assemblymember Rebecca Bauer-Kahan, Assembly Bill 1018 aims to impose strict oversight on automated decision systems (ADS) to prevent discrimination in the workplace and beyond.
If passed, AB 1018 would represent California’s most comprehensive attempt to regulate AI bias, placing significant compliance burdens on employers utilizing AI-driven hiring and workforce management tools.
Key Aspects of the Proposed Bill
AB 1018 introduces a robust framework governing AI-powered decision-making tools. Below are the most critical components for employers:
1. Opt-Out Rights and Human Review Requirements
Employers must provide employees and job applicants the opportunity to opt out of ADS-driven decisions in certain cases. Additionally, they must disclose whether human review occurs prior to finalizing an AI-driven decision.
2. Transparency and Appeal Rights
Employers are required to provide disclosures about AI-driven decisions, allow candidates and employees to opt out in specific situations, and establish an appeals process for adverse outcomes.
3. Expanded Definition of ‘Consequential Decisions’
The bill extends beyond hiring and termination to include wages, benefits, scheduling, promotions, performance evaluations, access to training, and workplace safety decisions, broadening compliance obligations for human resources functions utilizing AI.
4. Obligations for Employers Who Modify AI Systems
Employers who modify an AI tool beyond its original scope may incur legal responsibilities typically assigned to AI developers, increasing compliance risks for companies that customize AI tools.
5. Mandatory Performance Evaluations
AI developers are mandated to conduct annual audits and disclose any potential disparate impacts of their systems.
6. Third-Party Audits for Employers
Companies using AI tools in hiring, promotions, and terminations must undergo independent audits if their AI systems affect 6,000 or more individuals over three years.
7. Strict Data Management and Retention Rules
Employers are required to retain unredacted AI-related documentation for 10 years, including audit results, performance evaluations, and employee appeals. Non-compliance could lead to significant penalties.
8. Compliance Officers and Internal Review Obligations
Companies deploying AI decision-making tools must designate at least one compliance officer responsible for overseeing adherence to the law and investigating AI-related complaints.
9. Attorney General Oversight and Reporting Mandates
Employers may be required to submit unredacted performance evaluations to the California Attorney General upon request, with these records exempt from public records requests to protect sensitive business data.
10. California Consumer Privacy Act (CCPA) Compliance Considerations
Businesses subject to the CCPA must comply with AI-related privacy regulations, with future rulemaking likely to expand employer obligations regarding AI and employee data.
11. Enforcement and Penalties
The state Attorney General, Civil Rights Department, and Labor Commissioner can impose penalties of up to $25,000 per violation.
Comparison with Other AI Regulations
AB 1018 follows in the footsteps of Colorado’s groundbreaking AI law, which imposes anti-bias requirements on AI developers and users. However, unlike Colorado’s law, which primarily targets AI developers, California’s bill places significant responsibilities on businesses using AI in employment decisions.
Other jurisdictions have adopted varying approaches:
- Illinois’s AI Video Interview Act regulates AI-driven hiring assessments but is much narrower in scope.
- The EU AI Act classifies AI systems by risk level and imposes stringent obligations on high-risk applications, such as employment-related AI.
- Virginia’s AI Bill, awaiting action by the governor, focuses more on consumer protections and transparency, lacking the broad employment-related mandates of AB 1018.
If passed, California’s law would stand as one of the most expansive AI regulatory frameworks in the country, potentially shaping future legislative efforts nationwide.
Next Steps: Understanding the California Legislative Process
AB 1018 has just been introduced and must navigate California’s lengthy legislative process. The bill will first go through policy and fiscal committees, including the Assembly Privacy Committee, where amendments are likely. If it progresses, it will move to the full Assembly, then the State Senate, and ultimately the Governor’s desk. AB 1018 could reach the governor by mid-September, assuming it gains sufficient support.
Potential Challenges Ahead
Previous attempts to regulate AI bias in California have faltered due to business opposition and concerns over compliance costs. Similar legislation stalled last year before reaching a final vote.
However, AB 1018 has been drafted with input from labor and consumer advocacy groups, potentially giving it stronger momentum. Business groups, particularly those in Silicon Valley, are expected to advocate for exemptions or modifications, making the final form of the bill uncertain.
Governor Newsom’s stance on AI regulation has been mixed. While he supports consumer privacy and tech oversight, he has also promoted innovation-friendly policies, leading him to veto other AI legislation previously. Whether he ultimately signs AB 1018 will depend on how well the final version balances regulation with business concerns.
What Should Employers Do Now?
Even if AB 1018 does not pass this session, its core principles are likely to influence future AI regulations. Employers should proactively:
- Assess AI Tools in Hiring and HR Decisions: Conduct internal audits to understand how AI systems impact hiring, promotions, and terminations.
- Implement AI Governance Policies: Establish clear AI governance frameworks, including bias mitigation protocols and transparency measures.
- Prepare for Potential Compliance Obligations: Work with legal and compliance teams to align existing AI usage with anticipated regulatory requirements.
- Monitor Legislative Developments: Stay informed about AB 1018’s progress and be prepared for future compliance shifts.
