OpenAI's forthcoming model, GPT 5.6, is set for an unconventional release, deviating from its predecessors. Rather than a public launch, the company intends to provide access exclusively to a curated group of close partners, a decision reportedly influenced by directives from the Trump administration, according to The Information.
During a recent internal meeting, CEO Sam Altman reportedly informed staff that the government would be "approving access customer by customer" throughout an initial preview period. Altman further stated that a successful limited release could pave the way for a general, broader public launch "a couple of weeks later."
This development suggests the Trump administration is exerting pressure on OpenAI to adopt a strategy already voluntarily implemented by Anthropic: maintaining strict confidentiality and limited access to its most advanced AI models.
The Information also revealed that beyond administrative review, OpenAI personnel have "worked closely" with government officials on the impending release. The specific agencies reportedly advocating for this restricted deployment include the Office of the National Cyber Director and the Office of Science and Technology Policy.
This push for federal oversight marks a shift for the Trump administration, which initially advocated a "hands-off" stance on artificial intelligence. Earlier this month, President Trump signed an executive order mandating that certain AI companies voluntarily submit their new models to the government for testing and evaluation prior to public dissemination.
Earlier this year, Anthropic generated considerable debate with its announcement that its novel frontier cyber model, Claude Mythos, would be accessible solely to a limited group of partners through an initiative dubbed Project Glasswing. Anthropic justified this by asserting the model's immense power, positing that its misuse could lead to significant harm. Since then, commentators have deliberated whether Anthropic's rationale is a strategic marketing ploy or a genuine effort to prevent the misuse of a potent technology, with the truth potentially lying somewhere in the middle.
While cybercriminals have long leveraged automated tools, the advent of generative AI has provided them with unprecedented digital capabilities. Large Language Models (LLMs) have demonstrated proficiency in generating malicious software and, in some instances, can even execute complete ransomware attacks independently.
The particular apprehension surrounding frontier cyber tools such as Mythos stems from their purported ability to identify and exploit software vulnerabilities with a speed unachievable by human analysts. Given that numerous software systems contain latent bugs serving as potential entry points into enterprise networks, this presents a clear and substantial challenge for organizations managing intricate software infrastructure. Nevertheless, as these models are not publicly accessible, accurately assessing the full extent of their threat remains challenging.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.