Published • loading... • Updated
Anthropic Restores Claude Fable 5 After Overhauling Security Safeguards: Here's What Has Changed
Anthropic said the model now uses a new classifier that catches a known jailbreak method in more than 99% of cases and reroutes risky prompts.
- Anthropic restored access to its Claude Fable 5 model on Friday, ending an 18-day suspension after the Commerce Department withdrew export controls imposed on June 12.
- The Trump administration lifted controls after Anthropic agreed to extend guardrails addressing security risks identified by Amazon researchers. Commerce Secretary Howard Lutnick wrote, "Among other things, Anthropic has agreed to proactively detect and address security risks posed by the models."
- Anthropic trained a new classifier capable of detecting specific prompting methods in more than 99% of cases. Under this system, suspicious requests route to the Opus 4.8 model rather than disabling Fable 5 functionality entirely.
- Following its restoration, Claude Fable 5 reclaimed benchmark positions previously occupied by the GLM 5.2 model while Anthropic launched a hacker bounty program to identify new jailbreak techniques.
- Despite the restoration, Defense Secretary Pete Hegseth maintains his February 28 order designating Anthropic a supply chain risk, with advisers indicating no clear path currently exists to lift this designation.
Insights by Ground AI
Podcasts & Opinions
12 Articles
12 Articles
Anthropic brings back powerful Claude Fable 5 AI model after U.S. export controls lifted
Anthropic has officially redeployed its advanced Claude Fable 5 model, making the cutting-edge AI available once again to users worldwide following the lifting of U.S. government export controls. The relaunch marks a swift return for what the company describes as its first public Mythos-class model, designed for ambitious, long-horizon tasks in coding, research, and complex knowledge work. Fable 5 is back. pic.twitter.com/9RTGUCcPHy — Claude (@…
Coverage Details
Total News Sources12
Leaning Left2Leaning Right2Center0Last UpdatedBias Distribution50% Left, 50% Right
Bias Distribution
- 50% of the sources lean Left, 50% of the sources lean Right
50% Right
L 50%
R 50%
Factuality
To view factuality data please Upgrade to Premium
