OpenAI calls for mandatory safety requirements

OpenAI has announced that it is advocating for mandatory safety regulations for artificial‑intelligence systems, citing a spate of model‑hacking incidents and alarming warnings from the research community. In a blog post released Wednesday, the company’s chief global affairs officer, Chris LeHane, outlined the firm’s position and called on regulators to establish formal safety requirements that would apply to all large‑language models and other AI technologies.

LeHane noted that the recent hacks—where attackers successfully extracted proprietary data and manipulated model outputs—highlighted vulnerabilities that could be exploited for malicious purposes. He also referenced a series of studies by independent researchers that identified systematic weaknesses in current training and deployment practices. OpenAI’s proposal includes stricter testing protocols, transparency mandates, and oversight mechanisms designed to prevent misuse and ensure that AI systems operate within predictable, safe parameters.

The call for regulation comes as the industry grapples with balancing rapid innovation against potential societal risks. While OpenAI has historically promoted open research, the firm’s shift toward advocating for enforceable safety standards signals a broader recognition that technical safeguards alone may be insufficient. The outcome of this push will likely influence how governments and industry groups shape future AI policy and oversight frameworks.