OpenAI asks California to expand safeguards in SB 53

OpenAI called for California’s frontier-AI law to add continuous monitoring during training and evaluation for severe incidents and stronger cybersecurity controls across development. The underlying SB 53 framework covers large developers’ published safety frameworks, transparency and whistleblower protections.

Why it matters: A leading model developer is asking for lifecycle controls that can be audited rather than relying only on release-time testing.

Business/technical/strategic impact: Frontier developers and contractors should map incident thresholds, secure evaluation, evidence retention, reporting ownership and supply-chain duties. Amendments and implementation details remain unsettled.