Anthropic Neutered Its Cyber-Weapon AI Claude Fable 5 for Public Release
Oh joy, another AI company built a digital super-weapon and decided only governments get to play with the spicy version. While Uncle Sam gets a fully automated zero-day hacker, the rest of us get a heavily lobotomized corporate chatbot.
Under the secretive project code-named Glasswing, developers built a model called Mythos that specializes in autonomously discovering zero-day vulnerabilities and executing complex cyberattacks.
Instead of letting this digital anarchist loose on the public internet, Anthropic decided to split the release by keeping the raw hacker model exclusive to governments while giving developers a heavily restricted version called Claude Fable 5. The untamed original is reserved for intelligence agencies and defense partners who surely have only the most peaceful intentions.
This public version retains the core reasoning and massive context window of its scary sibling, but it comes pre-packaged with strict system blocks preventing it from actually writing malware or hacking servers. This compromise of a model is scheduled to debut on June 10 at the Code with Claude event in Tokyo.
It is reassuring to know that while the public gets an AI that politely refuses to write a slightly spicy email, government agencies will have an autonomous cyber-warfare machine ready to go. Surely nothing could possibly go wrong with this double-standard approach to AI safety.
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.