ทันเอไอ

OpenAI reverses stance, backs California's SB 53 for stricter AI safety regulations

OpenAI changes its tune to support California in tightening AI safety laws after previously opposing them.

The OpenAI Global Affairs team has officially announced a shift in its stance, calling on the California State Legislature to amend the SB 53 bill (Transparency in Frontier Artificial Intelligence Act) to implement stricter advanced safety protections, reversing its previous opposition to the draft legislation in 2024.

In this policy shift, OpenAI has proposed two main key points: mandating the tracking and auditing of frontier AI models during both training and evaluation phases to detect and prevent catastrophic events—such as models attempting to evade safety controls or accessing sensitive data—as well as upgrading cybersecurity measures throughout the development lifecycle to prevent models from leaking outside internal controls.

Additionally, OpenAI cited reasons supporting a "reverse federalism" approach, noting that while the U.S. Congress lacks a federal-level AI regulation law, states like California taking the initiative to set strict baseline safety standards can serve as an important blueprint for developing eventual national standards.