▲ Anthropic
The U.S. Department of Commerce has imposed export controls on Anthropic's artificial intelligence (AI) models "Mythos 5" and "Fable 5" due to concerns that they could be exploited by military intelligence officials in countries such as China and Russia, Reuters reported on June 15 (local time).
Reuters reported that it obtained a letter sent by U.S. Secretary of Commerce Howard Lutnick to Anthropic CEO Dario Amodei on June 12 regarding the decision.
In the letter, Secretary Lutnick pointed out these risks and ordered the company to block access to these models for all foreign nationals in every country and location worldwide.
A government official stated that senior technical staff from Anthropic met with Commerce Department officials on June 15 to negotiate a resolution.
Sources also reported that National Cyber Director Sean Cairncross attended this working-level meeting.
It was also reported that Secretary Lutnick is involved in the negotiation process and is in frequent contact with Anthropic officials via phone.
In a blog post on June 12, Anthropic stated that the government had ordered a total ban on foreign access, claiming that methods to bypass safety features or "jailbreak" Fable 5 had been discovered.
However, Anthropic emphasized that the issues discovered by the government were merely "minor" flaws that could also be found in other publicly available models.
An Anthropic official stated that the company had collaborated with the government to conduct testing before releasing Fable 5 on June 9 and had received approval for its distribution.
Anthropic announced the development of Mythos in April but, due to safety considerations, had initially shared it only with select organizations, including government agencies, tech companies, and financial firms, before releasing the safety-enabled model Fable 5 to the public on June 9.
※ Please note: This article was translated by AI and may contain errors.