Anthropic release Claude Mythos to the public despite 'risks' of super powerful AI
In a move that could change how people deal with cybersecurity threats, Anthropic has just released Claude Mythos to the public with a new name, Claude Fable 5, offering everyone the opportunity to utilize the game-changing tech despite the potential for risk.
The world of cybersecurity is constantly in a game of cat and mouse between those trying to find exploits and those looking to cover them up, and Anthropic's brand new AI tool has the potential to help both sides.
Starting out under the umbrella of 'Project Glasswing', Anthropic created an isolated test that certain companies – including tech giants like Microsoft and even government organizations – could take advantage of, with Claude Mythos being the model at the heart of it all.
What was once limited access has now been opened up to the public, bringing the capabilities of Mythos to all in a move that some have expressed concerns over.
What is Claude Mythos?
Claude Mythos is arguably the world's most powerful cybersecurity tool, as it's able to identify zero-day exploits and hacking vulnerabilities at a rate that was previously not deemed to be possible.
Advert
Anthropic has revealed that over 10,000 critical security flaws have already been identified within the systems of companies and organizations using Mythos, as the model works autonomously to find cracks that malicious actors would attempt to break through.
This is especially important within political organizations that might find themselves targeted by foreign cyber warfare, yet it can be useful to pretty much everyone — hence the desirability for Fable 5, the public version of Mythos, to be released.
How is Fable 5 different to Mythos?
Following it's launch this week, it is immediately clear that Fable 5 offers a significantly limited adaptation of the Mythos model that companies have been testing for a while now, as Anthropic is wary to give the public full access to the tool.
As reported by the BBC, Anthropic has outlined that safeguards and user limitations have been implemented within the design of Fable 5, yet its "capabilities exceed those of any model we've ever made generally available."
Its application isn't just limited to cybersecurity either, as it offers 'state-of-the-art' performance across "software engineering, knowledge work, vision, scientific research, and many other areas," according to the launch statement.
Mythos is still available for a "small group of cyberdefenders and infrastructure providers," as this enhanced version lifts certain safeguards to provide the "strongest cybersecurity capabilities of any model in the world."
What are the risks associated with Claude Fable 5?
What unfortunately arrives alongside the immense potential of a model like Fable 5 is almost equal risks, as this much power can potentially be used be malicious actors to cause significant harm.
OpenAI CEO Sam Altman has previously outlined the imminent risk of a 'world-shaking' cyberattack this year, and open source AI models will be the tool that allows this to happen.
"Releasing a model this capable comes with risks," Anthropic outlines. "Without safeguards, Fable 5's capabilities in areas like cybersecurity could be misused to cause serious damage."
There's no better way of finding out how to bypass the world's best cybersecurity tool than using the tool itself, as not only could criminals use this to find exploits before they're patched, but also find out the limitations of such a tool and focus on those areas.
That's why the safeguards implemented by Anthropic are vital to its safety, as certain queries will be deferred to the 'next-most-capable model', which in this case is Claude Opus 4.8.