ChatGPT is getting an upgrade. OpenAI’s GPT-5.6 will launch on Thursday, following a one month delay requested by the US government over concerns about cyberattacks.
The Trump administration had initially opposed any regulatory measures for artificial intelligence, but changed its stance after Anthropic unveiled Mythos, a model with advanced hacking capabilities that exposed thousands of security vulnerabilities in operating systems, popular software, and US government systems.
Worried that China and Russia could use American technology to design cyberattacks, the government is now requiring AI companies to submit their models for review before release.
The green light for OpenAI’s GPT-5.6 came after additional safety testing and meetings between OpenAI and government officials, according to Axios.
Anthropic faced a similar delay
Rival company Anthropic had previously been forced to withdraw its Mythos and Fable models after the government decided to impose export restrictions on the technology. Both models were relaunched last week after additional safety mechanisms were built in.
Until now, GPT-5.6 had only been available to select OpenAI partners whose information had been shared with US authorities.
The company announced the new model will be released in three versions: the flagship Sol, and two lower cost versions, Terra and Luna.
According to OpenAI, Sol matched Mythos level performance on ExploitBench, a benchmark for cybersecurity issues, and reportedly offers improved performance in biology, coding, and cybersecurity.
In China, Reuters reports, authorities are considering restricting foreign countries’ access to the country’s most advanced Chinese models.
