Anthropic's bug-hunting Mythos AI to be built into Cisco and IBM security tools

Anthropic's Mythos, an AI model known for its ability to find security flaws in software, is being built into security products from Cisco and IBM, giving many more organisations access to its bug-hunting abilities without handing them the model itself.

The deals were reported by The Register on 10 November. They mark a significant widening of access to a model that Anthropic has so far made available only through tightly controlled programmes, because in the wrong hands it could help attackers exploit software weaknesses.

Cisco: red teaming "super-charged"

Cisco announced at an event in Melbourne that the red-teaming tools in its AI Defense suite, which look for weaknesses and inappropriate activity in AI models, agents and the applications they power, will now be able to call on Mythos to test customers' set-ups.

"At a very high level, AI Defense has the ability to look at models and break them," said Jeff Schultz, Cisco's senior vice president for portfolio strategy and marketing. He said the tool was now "super-charged by Mythos". Cisco will make it available under customers' existing licences at no extra cost.

Natalie Mead, Anthropic's head of applied AI for Asia-Pacific, said there was a long queue of organisations wanting to use Mythos, and working with Cisco would make it easier to meet that demand.

IBM: finding what traditional scanners miss

A few hours later, IBM announced that it will also integrate Mythos into its Concert tool, which uses AI to scan applications for risks and weaknesses. IBM said it would use the model "to uncover weaknesses traditional scanners miss, determine which pose real risk and generate fixes for developer approval."

Access without handing over the model

Anthropic first made Mythos available through Project Glasswing, a programme that gave vetted organisations access to test their own code, and has since expanded access through other controlled schemes. The Cisco and IBM arrangements expose what Mythos can do to end users, while the model itself stays out of their hands.

On Monday, Anthropic also released version 5.2 of Mythos and its Fable model, which it says produce output up to 30% faster than the previous version at a lower price. The company said it had run "extensive prerelease safety testing" and concluded that "high-stakes risks from misalignment remain low".

Why it matters

Software vulnerabilities are at the root of many cyber attacks, from ransomware that shuts down hospitals and councils to breaches that expose customers' data. Tools that find and fix weaknesses before attackers do are among the most practical uses of AI, and making them available through products that companies already use could help smaller IT teams, not just tech giants.

For Birmingham's businesses, public bodies and its growing cyber security sector, the message is that AI is increasingly being put to work on defence. As ever, the best protection remains the basics: keeping software up to date, using strong authentication and having a plan for when things go wrong.