On Monday, Anthropic endorsed California’s pioneering bill, SB 53, proposed by state senator Scott Wiener. This proposed legislation seeks to impose unprecedented transparency requirements on major AI model developers such as OpenAI, Anthropic, Google, and xAI. Despite facing opposition from major tech groups and the federal government, Anthropic’s support represents a significant victory for the bill.
SB 53 aims to address frontier AI safety by mandating developers to establish safety frameworks and release public safety and security reports prior to deploying advanced AI models. It also includes whistleblower protections for employees reporting safety concerns. The bill focuses on mitigating catastrophic risks, such as AI models contributing to the creation of biological weapons or being utilized in cyberattacks, rather than more immediate issues like deepfakes.
While some argue that AI governance should be left to federal authorities, proponents of SB 53, including Anthropic’s co-founder Jack Clark, emphasize the urgency of establishing a state-level framework in the absence of federal standards. Despite the challenges, policy experts believe SB 53 is more restrained compared to previous AI safety bills and has a significant chance of becoming law.
California’s Senate has already approved a prior version of SB 53, although it awaits a final vote before reaching Governor Gavin Newsom’s desk. The bill specifically targets large AI companies with revenues exceeding $500 million, aiming to set state law requirements with financial penalties for non-compliance.
Ultimately, SB 53 could serve as a blueprint for future AI governance, balancing innovation with safety and setting a precedent for other states to follow.
