The US Department of Commerce has lifted export restrictions on the latest AI models from Anthropic after more than two weeks, the US tech company said on Tuesday.
Anthropic said it would begin restoring access to its Mythos 5 and Fable 5 models on Wednesday.
The US government had ordered the company to suspend access to the AI models for foreign nationals, including Anthropic employees. Anthropic then blocked both models entirely to comply with the requirement.
The artificial intelligence behind Mythos is particularly effective at detecting software vulnerabilities, including some that may have gone unnoticed for decades. US authorities and selected companies have previously used the technology to close such security gaps.
From the beginning, however, there were concerns that such AI could become a dangerous cyberweapon in the wrong hands.
Fable 5, released in mid-June, is based on Mythos technology, but its cybersecurity and biotechnology capabilities are blocked. Mythos 5, meanwhile, is the full non-public version and is intended to be available only to authorities and selected companies seeking to identify flaws in their own systems.
The restrictions also caused concern in Europe. Following the move against Anthropic, ChatGPT developer OpenAI granted access to its latest AI model only to partners approved in coordination with the US government.
The government's action was triggered by warnings that safeguards designed to prevent Fable 5 from being used to search for software vulnerabilities could be bypassed. Anthropic denied there was any danger. The government and the company held difficult talks over the recent weeks.
Commerce Secretary Howard Lutnick, in a letter published by US media announcing the lifting of the restrictions, said the company had agreed, among other things, to proactively identify and address security risks in its models.
