The mighty Claude Mythos 5 makes its public release as Fable 5

By now, you’ve almost certainly seen at least one AI demo that left you a little off: “That’s not only good; it’s terrifyingly “Whether we’re talking about AI stealing jobs from artists or generating novel proofs for mathematical mysteries, we’re constantly finding new ways in which AI-powered systems deliver results that make them seem almost uncomfortably powerful. We’ve heard recently anthropic sharing his concerns about the launch of his latest Claude Mythos model, which is just a little too good to find software vulnerabilities. Today, however, Anthropic is beginning to move forward with just that.

Automated bug finding is nothing new and computer scientists have been using tools like fuzzers for years, flooding the software with random input in hopes of causing a problem. But AI represents a much more potent threat, and as vulnerability hunters like Mythos have evolved, their creators have become appropriately wary of sharing their most powerful models.

Instead of letting its technology sit collecting dust forever, today Anthropic share the commitment How to implement Mythos responsibly and safely has been resolved. The key to this approach is to split things into two separate models: Claude Fable 5 and Claude Mythos 5.

Fable 5 is the model for the general public and is a full-featured solution for tackling analytical tasks. This doesn’t have to be limited to finding software vulnerabilities: Fable 5 can help with coding, has powerful vision analysis tools, and can even develop internal strategies over time. According to Anthropic, “Fable 5’s capabilities exceed those of any model we have made available to the public.”

But Anthropic also releases it with some conditions. Fable 5 is designed to resist attempts to use it to find the latest zero-day exploit, ready to be exploited by bad actors and wreak havoc on global computer systems. When someone tries to push Fable 5 outside of those limits, the model will fall again Close Job 4.8instead.

But then there’s Mythos 5, which internally is the same as Fable 5, but lacks many of those extra safeguards. The key here is that Anthropic plans to release it very selectively, inviting only trusted members of the cybersecurity community to use it. To get ahead of the bad guys, it’s important that a tool like this is used to identify bugs and fix them, and this fix is ​​intended to mitigate the risks of Mythos 5 being repurposed to be nasty.

Anthropic has developed a series of “classifiers” that try to recognize what users are trying to do with Fable 5, and not just try to block piracy. They are also designed to prevent anyone from using the model to develop ways to synthesize dangerous chemical or biological compounds, or to extract enough internal information to build their own Fable 5, without safeguards in place.

Hopefully this will be enough to keep this Mythos-based public version running safely, because we’re sure a lot of people will give it a try. very It is difficult to unlock all its capabilities.