OpenAI Now Supports California AI Safety Bill SB 53

OpenAI's Reversal on SB 53

In a striking policy pivot, OpenAI has reversed its stance on California’s Senate Bill 53, now urging lawmakers to strengthen the very AI safety legislation it previously opposed. The company’s shift comes after months of public criticism and internal debate over its initial lobbying against the bill, which would impose new transparency and risk-testing requirements on large AI models. OpenAI’s updated position, delivered in a letter to California legislators, calls for expanded provisions covering model deployment and incident reporting, arguing that “robust, enforceable standards are essential for public trust.” The reversal aligns with CEO Sam Altman’s recent congressional testimony, where he acknowledged that voluntary commitments alone are insufficient. Notably, the company still stops short of endorsing the bill’s most contentious clause—a private right of action for harmed parties—but its endorsement of stricter oversight marks a significant departure from its earlier “unworkable” characterization. This change may reflect growing pressure from employees and advocacy groups, as well as a strategic effort to shape final language before the legislative session ends.

What SB 53 Includes

SB 53 introduces a two-pronged approach to AI oversight, focusing on both pre-release evaluation and post-deployment accountability. The bill mandates safety testing for large AI models, requiring developers to conduct rigorous risk assessments before these systems are made available to the public. This testing is designed to identify potential harms related to bias, misinformation, and other high-risk failure modes.

Beyond the technical requirements, the legislation establishes critical whistleblower protections for employees. These provisions shield workers who report safety violations or other non-compliance issues from retaliation, ensuring that internal concerns can be raised without fear of job loss or professional repercussions. This creates a direct channel for accountability, empowering those with the deepest technical knowledge to act as a check on corporate practices.

Together, these measures aim to create a safer AI ecosystem by combining external verification with internal oversight.

Why OpenAI Changed Its Position

OpenAI’s reversal on SB 53 reflects a broader recalibration within the company, driven by growing concerns about advanced AI risks and a recognized need for regulatory clarity. The company has long warned about catastrophic misuse and loss of control, but its previous stance resisted binding state-level rules. Now, OpenAI argues that a clear legal framework—even one it initially opposed—can reduce uncertainty for developers and the public alike.

Key factors behind the shift include:

• Escalating real-world incidents involving AI-generated disinformation and deepfakes, which have intensified pressure on firms to act.
• A desire to preempt a patchwork of conflicting state laws by supporting a single, enforceable standard.
• Internal research indicating that voluntary safeguards are insufficient without external oversight.

OpenAI also emphasized that explicit rules would help define liability, creating a stable environment for responsible innovation. As CEO Sam Altman noted, “We need a framework that aligns incentives with safety,” a statement that marks a clear departure from earlier opposition.

The Impact of OpenAI's Support

OpenAI’s endorsement could be a decisive factor in the bill’s passage, lending significant credibility to SB 53 in the California legislature. As a leading AI developer, the company’s support signals to lawmakers that the proposed safety measures are both technically feasible and commercially viable. This backing may help overcome opposition from other tech firms that argue the regulations are too burdensome.

Beyond the immediate vote, OpenAI’s position is likely to shape future AI regulation in California and beyond. By endorsing the bill, the company sets a precedent for other states considering similar legislation, potentially creating a blueprint for national rules. The endorsement also pressures other major AI labs to publicly align with safety-focused governance or risk appearing resistant to oversight. If SB 53 becomes law, OpenAI’s support could encourage other jurisdictions to adopt comparable requirements, establishing California as a template for global AI policy. Conversely, if the bill fails, the endorsement still frames OpenAI as a proactive partner in regulation, influencing how future proposals are drafted and debated. Ultimately, the company’s stance may redefine the relationship between AI developers and state regulators.