california lawmakers pass ai safety bill sb 53 — but newsom could still veto

California AI Safety Bill Receives Senate Approval

California’s state senate has granted final approval to a significant AI safety bill. This legislation establishes new transparency requirements for major technology companies operating within the state.

Bill Details: SB 53

State senator Scott Wiener, the bill’s author, describes SB 53 as mandating transparency regarding safety protocols from large AI laboratories. It also incorporates protections for whistleblowers employed within these labs.

Furthermore, the bill proposes the creation of a public cloud resource, known as CalCompute, designed to broaden access to computational power.

Governor Newsom's Role

The bill is now awaiting the signature or veto of California governor Gavin Newsom. He has not yet publicly commented on SB 53.

Last year, Newsom vetoed a previous, more extensive safety bill also introduced by Senator Wiener. However, he did approve more focused legislation addressing concerns like deepfakes.

Previous Concerns and Revisions

At the time of the previous veto, Newsom recognized the necessity of safeguarding the public from the potential risks posed by AI technology.

He criticized the earlier bill for imposing “stringent standards” on large AI models without considering whether they were utilized in high-risk scenarios, involved critical decision-making processes, or handled sensitive data.

Senator Wiener indicated that the current bill incorporates recommendations stemming from a policy panel of AI experts convened by Newsom following the previous veto.

Revenue-Based Disclosure Requirements

Recent amendments to SB 53 introduce a tiered disclosure system based on company revenue.

Companies developing “frontier” AI models with annual revenues under $500 million will only be required to disclose general safety information. Those exceeding this revenue threshold will need to submit more comprehensive reports.

Industry Opposition

Several Silicon Valley companies, venture capital firms, and lobbying groups have voiced criticism of the bill and broader state-level AI regulation efforts.

OpenAI, in a recent communication to Newsom, argued for alignment with federal or European standards to avoid “duplication and inconsistencies” in safety regulations.

Constitutional Concerns Raised

Andreessen Horowitz’s head of AI policy and chief legal officer has asserted that certain state AI bills, including proposals in California and New York, may overstep constitutional boundaries.

This claim centers on potential violations of limits on state regulation of interstate commerce.

Previously, Andreessen Horowitz’s co-founders cited tech regulation as a contributing factor in their support for Donald Trump’s presidential campaign. The Trump administration subsequently advocated for a 10-year moratorium on state AI regulation.

Support from Anthropic

Anthropic has publicly expressed support for SB 53.

Jack Clark, Anthropic’s co-founder, stated in a post that while a federal standard is preferred, this bill “creates a solid blueprint for AI governance that cannot be ignored” in its absence.

SB 53 represents a significant step towards establishing a framework for responsible AI development and deployment within California.