OpenAI is lobbying California lawmakers to tighten provisions in recently enacted artificial intelligence safety legislation governing model deployment and computational thresholds.
Less than a year after helping shepherd California’s new artificial intelligence safety legislation through the state legislature, OpenAI is urging lawmakers to strengthen it. In a recent LinkedIn post published by its Global Affairs team, the company outlined its case for amending SB 53, the Transparency in Frontier Artificial Intelligence Act, which Governor Gavin Newsom signed into law in September 2025. OpenAI argues the statute requires additional enforcement mechanisms, specifically mandating that frontier AI models undergo monitoring during training and evaluation to detect attempts to bypass third-party security controls or access unauthorized confidential information.
Beyond targeted monitoring, the company is advocating for enhanced cybersecurity protocols throughout the entire model development lifecycle. These measures aim to prevent AI systems from circumventing their own internal safeguards, ensuring that security remains robust from initial design through deployment.
SB 53 was already a landmark piece of regulation upon passage. According to the Governor’s office, the law compels large frontier AI developers to publish comprehensive safety frameworks, establishes a formal reporting channel for critical safety incidents directed to California’s Office of Emergency Services, and includes whistleblower protections. It also empowers the state Attorney General to impose civil penalties on noncompliant companies. Additionally, the legislation created CalCompute, a public computing consortium designed to accelerate safety and equity research, and mandates annual legislative reviews to adapt to rapid technological shifts. At the time, Newsom framed the bill as proof that California could safeguard the public without stifling the AI sector, which has become a cornerstone of the state’s economy. Industry observers also characterized the law as a necessary stopgap amid congressional gridlock over federal AI policy.
This regulatory momentum aligns with what OpenAI describes as “reverse federalism”—a strategy wherein states pioneer compatible safety standards that can eventually scale into national policy while Congress remains inactive on comprehensive AI legislation. The company deliberately positioned its proposals as forward-looking rather than reactive, emphasizing that the goal is not “to write rules for one particular incident” but to construct a proactive framework that identifies vulnerabilities earlier and disseminates lessons across the industry. While these recommendations are widely viewed as constructive, they also ensure OpenAI retains a central role in shaping the very standards it seeks to implement.
Disclosure: In April 2025, Ziff Davis, Mashable’s parent company, filed a lawsuit against OpenAI alleging copyright infringement related to the training and operation of its AI systems.
Topics: Artificial Intelligence | Government | OpenAI
About the Author: Chance Townsend serves as General Assignments Editor at Mashable, covering technology, video games, dating applications, digital culture, and emerging trends. He holds a Master’s degree in Journalism from the University of North Texas and is a proud owner of an orange cat. His work has also appeared in PC Magazine and Mother Jones. Outside of journalism, Townsend enjoys cooking, prioritizing rest, and following Detroit sports. For tips or discussions regarding the Lions, he can be reached via Bluesky (@offbrandchance.bsky.social) or email at [email protected].