White House asks OpenAI to limit early release of GPT 5.6 to small group
.webp)
The White House has asked OpenAI to release its next artificial intelligence (AI) model, GPT-5.6, only to a small group of government-approved partners before making it available more broadly, according to a report by Axios.
The request was made by the White House's Office of the National Cyber Director and the Office of Science and Technology Policy as the Trump administration develops a framework to test and evaluate the security of advanced AI models.
This is also the first instance of the US government seeking to limit the rollout of a frontier AI model before its public release.
According to reports, OpenAI has been working with the Trump administration on the rollout of GPT-5.6. However, OpenAI CEO Sam Altman told employees that such a restricted release is "not our preferred long-term model" and that the company would work towards a more sustainable approach for future launches, the Axios report said.
Why is the Trump administration limiting GPT-5.6’s rollout?
The Trump administration is not taking a heavy-handed approach, the report said, adding that it intervened because the new model will have “Mythos-like” capability.
The models are so powerful that the US government wants to be sure companies have adequate safeguards in place, the report added.
As part of this effort, the White House has reportedly been briefed on GPT-5.6's capabilities, while Commerce Secretary Howard Lutnick discussed the model with Altman and sought to ensure that relevant government agencies had tested and approved it before a broader rollout.
The move also follows President Donald Trump's executive order earlier this month directing federal agencies to establish a voluntary testing protocol for frontier AI models before their release. According to the report, debates within the administration over how restrictive the framework should be had delayed the order for weeks.
What does Mythos-like capability mean?
The reference to "Mythos-like" capability comes from Anthropic's Mythos AI model, which was unveiled in April. The model was designed to identify cybersecurity vulnerabilities and potential threats before they can be exploited by hackers.
However, Anthropic chose not to release the model publicly, saying it was too powerful for unrestricted access.
According to the company, Mythos can identify previously unknown flaws in IT systems that could potentially be exploited by cybercriminals.
Instead of a public launch, Anthropic initially made the model available only to a limited group of technology companies under Project Glasswing, an initiative the company described as an effort to secure the world's most critical software.
Other Latest News
.webp)
A subsidiary of UK-based Vedanta Resources has accepted bids worth $1.75 billion for a three-tranche issuance of dollar bonds, as it seeks to refinance high-yielding outstanding debt, three merchant bankers said on Friday.
Vedanta Resources Finance II has raised $500 million through six-year bonds at a coupon of 7.00%, $700 million through eight-year bonds at a 7.3750% coupon, and $550 million through 11-year bonds at a coupon of 7.75%, the bankers said.
.webp)
The death toll is rising in Venezuela after back-to-back earthquakes struck west of the capital on Wednesday, toppling buildings and severely damaging the country’s main international airport.
The natural disaster has killed at least 188 people and injured more than 1,500 others in a country where emergency response capacity is limited after years of political and economic turmoil.
.webp)
A weak monsoon is emerging as India’s next inflation risk, with a developing El Niño threatening to curb rainfall and push up food prices just as an easing oil relieves broader price pressures.
The southwest monsoon, which accounts for about 70% of India’s annual rainfall, is vital for the $300 billion farm economy and has an outsized influence on food prices, rural demand and broader economic output.
.webp)
Hon Hai Precision Industry Co. has said that its subsidiary, Foxconn Singapore, acquired around 351.73 million common shares in its India subsidiary, Foxconn Hon Hai Technology India Mega Development Ltd, for $37.2 million, according to a stock exchange filing.
The Taiwanese electronics giant is the world's largest manufacturer of iPhones.
.webp)
In a significant relief to manufacturers of products like toys, personal protective equipment, air conditioners and footwear, the government on Thursday allowed an alternative way to ensure quality standards without the rigours of following the quality control orders (QCOs) applicable to those sectors, in a notification issued late on Thursday.
.webp)
.webp)
.webp)
.webp)
.webp)
