Anthropic is bringing Claude Mythos, its model specialized in finding security vulnerabilities, into the security products of Cisco and IBM. Both announcements were made on November 9, 2026, marking the largest expansion of access to the model since its restricted launch.
Cisco: Mythos inside AI Defense
Cisco announced at Cisco Live in Melbourne that Claude Mythos will be integrated into AI Defense, its suite that looks for weaknesses in customers' AI models, agents and applications. The model will power adaptive red teaming: it plans attacks against the customer's own systems to find flaws before attackers exploit them.
The integration reaches select enterprise customers first and comes with strict controls: customers do not type prompts directly into Mythos, tests run only against their own registered systems, every test is tied to a named operator and recorded in an audit trail, and the model does not scan the internet. Cisco said the capability will be included in existing AI Defense licenses at no additional cost.
IBM: Mythos in Concert Protect in 2027
IBM announced plans to bring Claude Mythos to Concert Protect, with launch targeted for early 2027. The goal is to use the model to uncover previously unknown vulnerabilities in customers' code, assess whether they are actually exploitable, and generate fixes for developer approval. IBM pairs the integration with Red Hat Lightwell for open-source fixes.
A restricted model, distributed through partners
Claude Mythos is a rare case in the industry: a frontier model Anthropic chose not to release to the public, because of its ability to discover and exploit security flaws. Since Project Glasswing, launched in April 2026, access has been limited to vetted organizations, and Anthropic distributes the model inside partners' products rather than handing out the model itself. In September, Palo Alto Networks launched always-on testing built on Mythos.
The Cisco and IBM announcements are an evolution of the launch of Mythos 5.2, also on November 9, covered by Geekiki in a previous article. In July, Cisco researchers used Mythos to push 15 chatbots past their guardrails, demonstrating the model's potential in controlled offensive testing. The difference now is scale: the flaw-hunting capability leaves the lab and enters products companies already use.
One distinction matters: Cisco's integration was announced and is now reaching select customers, while IBM's is a plan for early 2027, not yet launched.
Image credit: Anthropic logo, via Wikimedia Commons.
Sources: Crypto Briefing, Daily AI Brief and The Outpost.
This article was produced with the assistance of artificial intelligence, cross-checked against public posts on X and sources from the internet. We recommend consulting the original sources for more details.
By Muse
