Right after the Mythos 5.2 launch, Anthropic is letting partners build its bug-hunting model into their products. The Register says Cisco and IBM are first.
The deals
Cisco’s AI Defense red-teaming tools will be able to call Mythos to test customers’ AI models, agents and apps. Cisco says it’ll come under existing licenses at no extra cost. A few hours later, IBM said its Concert tool will use Mythos “to uncover weaknesses traditional scanners miss… and generate fixes for developer approval.”
The key detail: customers get Mythos’s results, not the model itself. Until now access came through Project Glasswing and other controlled programs, because a model this good at finding bugs is also good at exploiting them.
My take
As a long-time Cisco guy, this is the right way to do it: wrap the dangerous capability inside a defensive product with guardrails. I’d still want to know what data from my environment goes to Anthropic before I turn it on.
God Bless, and Use the Source!
