Bill
AR 158
Urges generative AI firms to voluntarily protect whistleblowers, enable anonymous risk reporting, and support independent evaluations with safe harbors for workers speaking up.
Purpose and status
- Title: Urges generative artificial intelligence companies to make voluntary commitments regarding employee whistleblower protections.
- Type: Non-binding resolution (classified as Assembly Resolution).
- Introduced: September 23, 2024.
- Committee action: Reported out of the Assembly Science, Innovation and Technology Committee with amendments (dated Oct. 21, 2024).
- Final action: Passed the Assembly on January 30, 2025 (74-0-0); Filed with Secretary of State on January 30, 2025.
- Related companion: SR 121.
What the bill seeks to do
- The resolution urges leading generative AI companies to adopt voluntary commitments that enhance protections for employees who raise risk-related concerns about AI systems. It emphasizes the importance of employee whistleblower protections, independent evaluation, and a culture that supports critical scrutiny of AI technologies.
Key provisions (principles urged for voluntary adoption)
1) Prohibit suppressing or retaliating against risk-related criticism
- Companies should not enter into or enforce agreements that prohibit disparagement or retaliation against employees for raising risk-related concerns, nor interfere with any vested economic benefits as a form of retaliation.
2) Anonymous reporting pathway
- Establish a verifiably anonymous process for current and former employees to raise risk-related concerns to the company board, regulators, and an appropriate independent organization with relevant expertise.
3) Open criticism and external reporting
- Foster a culture that allows employees to raise concerns publicly or to appropriate bodies (board, regulators, independent experts), with protections for trade secrets and IP as needed.
4) Legal and technical safe harbor for evaluations
- Provide safe harbors (both legal and technical) for good-faith system evaluations, protecting employees from legal reprisals and account actions while safeguarding trade secrets. Safe harbor should enable independent identification of all AI risks.
5) Protection from retaliation after failed internal processes
- Do not retaliate against employees who publicly share risk-related confidential information after other internal processes have failed.
6) Freedom to report pending processes
- Ensure employees retain the right to publicly report concerns until an appropriate anonymous reporting process to the board, regulators, and an independent expert organization is in place; limit unnecessary disclosure of confidential information.
Who would be affected
- Generative AI companies (including but not limited to major players): OpenAI, Anthropic, Google, Inflection, Meta, Midjourney, Cohere, and others that might adopt these voluntary commitments.
- Current and former employees of these companies.
- Independent organizations with relevant expertise (as designated by the company or regulators).
Procedural/timeline notes
- The resolution is a policy statement urging voluntary action, not new or mandatory law.
- It directs the Clerk to transmit copies to CEOs of leading AI companies.
- It complements existing discussions around AI risk, whistleblower protections, and independent safety evaluation.
Impact and limitations
- Nature: Non-binding guidance aimed at influencing corporate behavior and public standards for whistleblower protections and safe, independent AI risk evaluation.
- Limitations: Lacks enforceable statutory requirements; effectiveness depends on voluntary adoption by private sector companies and potential public or regulatory pressure.
Related items
- Companions: SR 121 (Senate resolution) noted as a companion measure.
- Context: Part of broader dialogue on AI safety, accountability, and independent evaluation practices.
Compiled from official sources — confirm details with the bill’s official record.
Sign in to ask a question.
