Share
Anthropic is letting two of enterprise tech’s oldest names plug its Mythos model into their security products.
Cisco and IBM will embed the bug-finding AI into tools their customers already use.
The pitch is simple. Find the holes in software before attackers do, and use an AI that is good at thinking like an attacker to do it.
What Cisco and IBM are actually building
Anthropic disclosed the expansion on November 9, 2026. Cisco is folding Mythos into its AI Defense product, where it will power adaptive red teaming of AI applications and agents.
Red teaming is the security world’s version of hiring a burglar to test your locks. The idea is to attack your own systems on purpose so you find the weak points first.
Jeff Schultz, Cisco’s senior vice president for portfolio strategy and marketing, described AI Defense as a tool built to probe models and break them. He said Mythos will help it surface more threats, which AI Defense can then work to fix.
“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.
“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.
IBM is taking a slightly different route. It plans to add Mythos to Concert Protect, a product aimed at discovering unknown vulnerabilities in customer code and validating whether they can actually be exploited.
IBM’s rollout is targeted for early 2027. So Cisco customers get the upgrade first, while IBM’s version is still a plan on the calendar rather than a shipped feature.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
We respect your privacy. Unsubscribe anytime.
A model Anthropic would rather not hand out freely
Mythos is not a general-purpose chatbot with a security hobby. Claude Mythos versions 5.1 and 5.2 are purpose-built for cybersecurity work.
Those models can identify software vulnerabilities and plan multi-step attacks. Anthropic has deliberately restricted that capability to approved partners.
That restraint traces back to Project Glasswing, which Anthropic launched in April 2026. The program gives vetted organizations access under strict security controls instead of opening the model up through a public API.
The guardrails are specific. Mythos only runs against registered targets, requires named-operator authentication, and keeps comprehensive audit logs.
For those who do get access, the meter runs at $10 per million input tokens and $50 per million output tokens. Output costs five times as much as input under that pricing.
Why the gatekeeping is the real story
Every security tool is, in some sense, dual-use. A model that can map out a multi-step attack for a defender can, in the wrong hands, map one out for an attacker.
Routing access through established vendors is one way to scale without throwing the doors open. Customers reach Mythos through products wrapped in Cisco’s and IBM’s own controls, rather than through raw API keys.
These integrations are part of a broader push to arm defenders with advanced AI before adversaries get comparable tools.
What this means for the security market
For enterprise buyers, the immediate change is practical. Teams already paying for AI Defense or planning around Concert Protect could get frontier-grade vulnerability hunting without signing a separate deal with Anthropic.
For Anthropic, the move turns a tightly held research capability into a revenue channel.
Watch three things from here. First, whether IBM hits its early 2027 window for Concert Protect. Second, whether Anthropic adds more partners to the Mythos roster or keeps the club this small.
