The Trump administration is expected to lift export controls on two of Anthropic’s most powerful artificial intelligence models on Tuesday evening after the company reached an agreement with the Commerce Department, according to a person familiar with the matter.
The department will lift restrictions on both the Fable 5 model and the more powerful Mythos 5 model, which has so far been approved for release only to select companies and government agencies, this person said.
The move is expected to be announced by Anthropic co-founder Tom Brown, sent by Commerce Secretary Howard Lutnick, who, along with national cybersecurity director Sean Cairncross, led the Trump administration’s efforts to resolve the dispute with Anthropic.
These changes come as Anthropic is working with the Department of Commerce and the White House, the people say, to strengthen safeguards against users bypassing Fable’s security restrictions to gain access to confined capabilities, particularly those related to cybersecurity.
Anthropic initially maintained that the administration’s fears of a prison break were exaggerated. The company said it cannot guarantee there will be no jailbreaks that could unlock more powerful capabilities of the confined Mythos model, which last week was approved for confined release to select companies and government agencies.
In recent weeks, Anthropic has changed tactics in trying to get Fable back online, which has also meant changing the company’s communication style with the administration. WIRE we previously reported that Anthropic CEO Dario Amodei had recently been replaced during meetings with co-founder Tom Brown, whom officials liked better on a personal level.
Anthropic also assured the administration that it would try to reduce prison escapes by creating more tough safeguards, effectively telling the administration what it wants to hear rather than revisiting the conceptual question of whether prison escapes can be stopped, the people said.