Not through a decree from President Trump, but through a letter from Secretary of Commerce Howard Lutnick, Anthropic can offer its Claude Mythos AI model again. Several US media reported this independently after gaining insight into the letter.
Read more after the ad
CNBC quotes the minister from the letter: “I have determined that appropriate security measures have been put in place to allow certain trusted partners to access the Claude Mythos 5 model.” In the original, Lutnick writes here about “safeguards”, a kind of guardrails that are intended to prohibit AI models from carrying out dangerous tasks after a jailbreak. Semafor and Politico also report on the letter.
Who is allowed to have the myth remains unclear
The US government views the modern LLMs from Anthropic and OpenAI, among others, as a possible threat to national security. Anthropic has already been classified as a “supply chain risk”. The company is therefore subject to strict export controls, such as those common to arms manufacturers. Therefore, even after the total lockdown is lifted, Mythos will not remain accessible to everyone.
According to Politico, there should only be a little more than 100 companies, literally “agencies”. There have been reports for some time that the NSA is supposedly allowed to use Mythos, even for offensive cyber attacks. There is no list of organizations that should have access to Mythos in the current reports. The Fable 5 model, which is supposed to be a defused version of Mythos, remains completely blocked for the time being.
The legal basis for the strict regulation of AI models from US providers remains unclear. Numerous other companies are supporting Anthropic in its lawsuits against its classification as a security risk. In a brief to the court as amicus curiae, they described the U.S. government’s actions as an “inappropriate and arbitrary use of power with serious consequences for our industry.”
Read more after the ad
(NO)
