US Government Influence Puts OpenAI's New AI Model on US-Only Preview

On Friday, OpenAI initiated a restricted, US-exclusive preview of its advanced AI model series for a select group of partners. This action, according to the company, was undertaken following a specific request from the United States government. This rollout occurs a fortnight after the White House surprised Silicon Valley by instructing Anthropic, a competitor to OpenAI, to prohibit all non-US citizens from using its Fable 5 and Mythos 5 models, citing concerns over national security. Anthropic promptly complied by discontinuing access to those particular models, stating it could not reliably enforce restrictions on foreign nationals.

The most recent models from prominent AI firms, such as Anthropic’s Mythos series and now OpenAI’s GPT-5.6, have generated considerable apprehension due to their reportedly unprecedented capacity to pinpoint software vulnerabilities – flaws in code that hackers could potentially exploit. Under increasing scrutiny regarding the novelty of their capabilities, former President Trump earlier this month signed an executive order establishing a voluntary federal review process for national security risks associated with advanced AI models before their public release.

The White House has provided minimal details on how it intends to enforce this executive order, which companies are understood to be participating in voluntarily, or which specific models would be subject to its review guidelines. This intervention was particularly striking given that the White House has generally advocated for reduced AI oversight, even attempting to prevent states from formulating their own regulations. The decisive action taken against Anthropic has attracted accusations of government overreach, and OpenAI itself expressed unease with the protocol it was mandated to follow for its new models.

OpenAI confirmed that it had briefed the US government on the capabilities of its new models prior to their launch. Furthermore, at the government’s insistence, it is commencing with a limited preview exclusively for a trusted group of US-based partners, whose identities have been shared with relevant authorities. However, OpenAI clarified that employees of these companies or entities working overseas would also be granted access to the new models.

“We don’t believe this kind of government access process should become the long-term default,” OpenAI stated in a blog post. “It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks.”

When Anthropic was initially targeted, some observers believed the safety-focused company was being unfairly singled out by the Trump administration for political motives. In a previous disagreement with the White House, Anthropic had reportedly angered Trump’s team by declining to permit its technology’s use for mass surveillance and autonomous weaponry, which led to the Pentagon canceling its contracts with the company. That dispute is currently being litigated across two separate lawsuits.

OpenAI’s GPT-5.6 series is composed of three distinct models: Sol, designated as the company’s new flagship; Terra, a mid-range model designed for daily operational tasks; and Luna, an economical and high-speed alternative. Upon wide availability, Terra is expected to be priced at half the cost of its predecessor, GPT-5.5, as OpenAI strives to secure its customer base amidst intense competition from Anthropic and Google. Both OpenAI and Anthropic have submitted confidential IPO documents to US regulators, aiming for public listings that could value them at nearly $1 trillion, significantly raising the commercial stakes in the ongoing AI arms race between them.