Tech
After spooking Trump into safety testing, Anthropic AI models get global release
The US government has lifted export restrictions on Anthropic's advanced Claude models, Fable 5 and Mythos 5, following a temporary ban enacted on June 12 due
Key takeaways
- The US has lifted export curbs on Anthropic's newest Claude models, Fable 5 and Mythos 5.
- Anthropic agreed to expand its partnership with the government to address national security risks.
- Mythos 5 can find and exploit software vulnerabilities more effectively than any other model.
- Fable 5's tightened safety safeguards may cause some benign prompts to be blocked during routine coding.
- The government and private sector have worked together in an unprecedented way.
The US government has lifted export restrictions on Anthropic's advanced Claude models, Fable 5 and Mythos 5, following a temporary ban enacted on June 12 due to national security concerns. Commerce Secretary Howard Lutnick confirmed that Anthropic no longer needs a license for exports or transfers of these models after coordinating with the government to address risks. Fable 5 is now available globally, and US organizations have had access restored to Mythos 5 since June 26. To secure this release, Anthropic expanded its government partnership, established a 24/7 jailbreak monitoring team, and launched a HackerOne security program. However, Anthropic noted that the tightened safety classifiers on Fable 5 may cause a trade-off where benign prompts are blocked during routine coding and debugging tasks.
In their words
“the government and private sector have worked together in a way we have never seen before and this foundation of America First is unprecedented.”
“no longer need a license for exports or in-country transfers of its Claude Mythos and Claude Fable AI models.”
By the numbers
- 99%
- Cases in which the reported jailbreak method is blocked
- 24/7
- Dedicated internal team monitoring reports of emerging jailbreak threats
How it unfolded
- Mythos release spooks Trump into requesting voluntary safety testing
- Commerce Department orders Anthropic to shut off advanced model access
- US organizations have access restored to Mythos 5
- US lifts export curbs; Fable 5 available globally
Why this matters
-
It is probably impossible to build an AI model fully impervious to jailbreaks.
-
The government is moving too slowly to establish formal AI safety standards.
-
The biggest question is how equivalently dangerous capabilities from China will be handled.
Turn stories like this into views
Ravenclip finds the Tech news, makes the video, and posts it before attention moves on.
Common questions
- What happened with Anthropic?
- The US has lifted export curbs on Anthropic's newest Claude models, Fable 5 and Mythos 5.
- Where can I read the original report?
- Read the full report at ars_technica.
