OpenAI will expand access to its new artificial intelligence (AI) model starting Thursday, with the White House saying it had no role in the decision.
• Also read: OpenAI deploys its new model under surveillance by the US government
• Also read: OpenAI launches GPT-5.5, its most powerful artificial intelligence model to date
“We are now extending preview access globally,” OpenAI said overnight from Tuesday to Wednesday on the social network X, without providing further details.
At the end of June, OpenAI accepted a very limited deployment of this new model, called GPT-5.6, with partners validated by the American government.
But American media Axios reported Tuesday evening that the United States had finally given the green light to an expanded launch of the model, following technical tests and meetings between government officials and company representatives.
Asked by AFP, the White House refuted having given the green light, “no authorization (being) required” to market a given model.
“The decision on the date and extent of a (model) release is entirely at the discretion of the companies,” assured a White House official, who referred to the provisions of the decree taken at the beginning of June by Donald Trump.
This text introduces the possibility for an AI player to have a model tested before marketing on a voluntary basis.
Contacted by AFP, OpenAI did not immediately respond.
Long self-proclaimed champion of deregulation, Donald Trump’s government abruptly changed course at the beginning of June, worried about the new capabilities of the most sophisticated AI programs in terms of cyberattack.
It began by prohibiting Anthropic, OpenAI’s great rival, from allowing non-US nationals to use its two cutting-edge models, Mythos 5 and Fable 5 (a restricted version of the first), citing risks to national security.
Saying it was unable to enforce this injunction, the Californian start-up simply withdrew the two models. At the end of June, the government finally lifted its ban and Anthropic brought Mythos and Fable back online.
The GPT‐5.6 range includes three versions with cosmic names: Sol, OpenAI’s new flagship model; Terra, intermediate version intended for everyday uses; and Luna, a faster and less expensive version.
Pulling
Once widely available, Terra will be offered at half the price of GPT‐5.5, OpenAI said, as the group seeks to build customer loyalty in an increasingly competitive market against Anthropic and Google.
The multiple twists and turns experienced by Mythos and GPT-5.6 demonstrate tensions within the Trump government between opponents of any form of control of AI and supporters of a reasoned approach, focused on the risks of this technology.
In addition to the arrival of much more powerful models, the turnaround of the Trump government also corresponds to the departure, in May, of David Sacks, AI referent at the White House and champion of deregulation of the sector.
Until now, no government has publicly required an AI laboratory to limit or suspend access to a model after its launch. The Chinese authorities set constraints on its local companies, but before their exit.
The American government’s interventions targeting Anthropic and OpenAI have stunned a large part of the industry, now concerned about access to the most cutting-edge American models.
For many observers, current developments are likely to favor open and inexpensive models, such as those of the Chinese DeepSeek, more attractive for clients keen to avoid any dependence on Washington.
The new posture of the American government could affect the valuation of OpenAI and Anthropic, which are both preparing their IPO in the medium term.
The two rivals have already filed their IPO on Wall Street and currently boast valuations close to a trillion dollars each.
