NJ AR158

Calls on generative AI companies (including OpenAI, Anthropic, Google, Meta, Midjourney, and Cohere) to voluntarily adopt whistleblower protection principles for employees raising risk-related concerns

Urges companies to not enforce agreements prohibiting criticism of the company for risk-related concerns and to not retaliate against employees who raise such concerns

Requests companies establish anonymous processes for current and former employees to report AI safety concerns to company boards, regulators, and independent organizations

Encourages companies to provide legal and technical safe harbor for good faith safety evaluations of AI systems, protecting researchers from legal action or account termination

Expresses concern that existing whistleblower protections are inadequate and that broad confidentiality agreements prevent AI employees from voicing safety concerns about technology risks including misinformation and inequality