Anthropic's Mythos model is heading into Cisco and IBM security products

[JeffHardy_Stunner]

Anthropic's bug finding model is going mainstream in enterprise security. The Register reports that Cisco announced in Melbourne that red teaming in its AI Defense suite can now call on Mythos, at no extra cost under existing licences

Cisco's Jeff Schultz, senior vice president for portfolio strategy and marketing, said AI Defense has the ability to look at models and break them, and called it super charged by Mythos. IBM will also add Mythos to its Concert product to uncover weaknesses traditional scanners miss and draft fixes for developers to approve

Natalie Mead, Anthropic's head of applied AI for Asia Pacific, spoke of a long queue of buyers, but would not say whether other deals exist. Access began through Project Glasswing, and customers get the model's functions rather than the model itself

Anthropic also says Fable 5.2 has its strongest ever protections against distillation attacks. Distillation is when someone tries to copy a model's abilities by training another model on its outputs

Finding security holes before attackers do is one of the most useful things AI can do. Building it into products that companies already use makes adoption much easier. Keeping humans in the loop to approve fixes is a sensible safeguard

Would you trust AI to find and fix security bugs? And how will this change the work of security teams?

Anthropicwww.theregister.com