The U.S. government orders Anthropic to shut down its most powerful Claude models, Fable 5 and Mythos 5
So Anthropic says now even some of its own employees, who built Anthropic’s most powerful new AI mod...
AI BriefingLast Friday, the U.S. government issued Anthropic an export-control directive requiring it to shut down its most powerful models, Fable 5 and Mythos 5. The move followed the discovery of a jailbreak method that could make the models provide cybersecurity assistance they should have refused. Commerce Secretary Howard Lutnick said the models would be subject to export restrictions for foreign nationals both outside and inside the United States until the national security system was strengthened, which could take several weeks. Anthropic responded that the jailbreak technique was narrow and involved only a small number of known vulnerabilities, and that other publicly available models could provide similar capabilities. However, because the company could not verify users’ nationalities in real time, it had to disable the models for everyone, including members of its own international teams. Anthropic also said that the industry currently cannot achieve perfect jailbreak resistance and that all safeguards are vulnerable to non-general-purpose jailbreaks.
Reported by 6 additional sources
- Who decides how dangerous AI is? US government restrictions on Anthropic Fable 5 spark controversyThe Verge: AI · June 18
- Explaining the political power struggle between the White House and Anthropic over the Fable modelsThe Verge: AI · June 18
- Inside Anthropic's clash with the government over Claude Mythos 5The Verge: AI · June 16
- Anthropic and the White House clash anew over the ban on Fable 5 and Mythos 5The Verge: AI · June 16
- Amazon security research reportedly led the White House to restrict Anthropic’s Fable 5 and Mythos 5The Verge: AI · June 14
- 👀 Hope Fable 5 and Mythos 5 return soonRohan Paul · June 14
Source: Rohan Paul · x.com