White House Urges OpenAI to Delay Public Release of GPT 5.6 Over Safety Fears
The U.S. government is asking OpenAI to initially release its new GPT 5.6 model only to select partners and approve each customer individually, mirroring Anthropic's previous approach. This follows an executive order on voluntary AI testing and concerns that powerful models could be misused for cyberattacks.
According to The Information, OpenAI's next-generation model, GPT 5.6, will not see a public launch like previous versions. CEO Sam Altman reportedly told employees that during a preview period, the government will approve access customer by customer. If the limited release goes well, OpenAI hopes to follow with a broader public release a couple of weeks later.
The move comes after pressure from the Trump administration, which initially positioned itself as taking a "hands-off" approach to AI but has recently pushed for federal oversight. The Information reports that the Office of the National Cyber Director and the Office of Science and Technology Policy requested the limited release. Altman also noted that OpenAI staff worked closely with the government on the upcoming release.
Earlier this month, Trump signed an executive order directing certain AI companies to voluntarily submit new models to the government for testing and evaluation before public release. This echoes Anthropic's decision earlier this year to release its Claude Mythos model only to a small group of partners through Project Glasswing, sparking debate over whether the move was marketing or legitimate safety concern.
Experts note that generative AI has given cybercriminals more digital ammunition. Large language models can write malware and even execute autonomous ransomware attacks. Frontier cyber tools in particular can identify and exploit software vulnerabilities faster than any human analyst. However, because these models remain closed to the public, assessing the true threat level is difficult.
