Administration to Decide Access to ChatGPT Upgrade

OpenAI's newest AI is here—but whether anybody can use it is up to the Trump administration. The company on Friday said its GPT-5.6 models will roll out first to a small set of users cleared by the administration, in what OpenAI called a temporary gatekeeping step that it argues shouldn't become standard practice and that the AP calls an unprecedented vetting. The White House has been pushing to tighten case-by-case controls on powerful AI systems in the name of national security, the Wall Street Journal reports.

GPT-5.6 will launch in three versions geared toward reasoning, routine tasks, and heavy-duty workloads, and has shown advances in autonomous task completion plus sensitive areas like biology and cybersecurity. Similar capabilities in Anthropic's Mythos line helped trigger stricter oversight, including a recent ban on foreign access to two of Anthropic's latest models that led the rival to shut them down entirely. OpenAI, which has generally aligned itself with President Trump's goal of topping China in the AI race, says GPT-5.6 is being reviewed by a Commerce Department testing unit as a new executive order brings more cyber and national security officials into the approval process. The company says it hopes to open access "in the coming weeks" and calls the current approval procedure a brief transition period.

There is no process for individual users to gain access to the new OpenAI model, per the Washington Post; only companies can receive government approval. Trump won the support of tech executives in his campaign last year by promising them greater freedom; they had complained that the Biden administration was too restrictive. But the administration recently has become more involved. "In a matter of weeks, US federal AI policy has gone from implausibly libertarian to increasingly draconian and opaque," Dean Ball, a former Trump AI adviser, posted Friday.