Cisco announced at the Cisco Live event in Melbourne that it will integrate Anthropic's Claude Mythos into AI Defense to help select enterprises test their AI apps and agents. According to The Next Web, this integration brings automated red teaming to enterprise customers looking to find system weak spots.
Key takeaways
- Cisco revealed the integration during its Cisco Live event in Melbourne.
- Claude Mythos will initially roll out to select enterprise customers without a firm launch date.
- The model plans tests to find flaws in AI systems before hackers can exploit them.
- Cisco researchers used the model in July to push 15 chatbots past their guardrails.
What is Claude Mythos used for in Cisco AI Defense?
Claude Mythos serves to plan attacks and generate testing ideas against AI systems in a practice known as red teaming, as reported by The Next Web. The model targets customer-owned AI apps and agents to uncover vulnerabilities before deployment. In July, researchers at Cisco used the model to bypass the guardrails of 15 different chatbots.
Users do not get direct open prompt access to the model. Instead, it operates exclusively on fixed jobs against registered systems. Also, every test is logged to a named person for audit purposes, and administrators can halt tests whenever necessary. The system cannot scan the wider internet, according to a blog post from Cisco.
Industry adoption of Anthropic models
Security firms are rapidly incorporating the model into their own software offerings. Palo Alto Networks introduced always-on testing using both Mythos and an OpenAI cyber model in September. Also, IBM announced plans to embed Mythos into its Concert Protect tool in early 2027.
Anthropic prefers embedding the model inside partner products rather than distributing it directly. The company released Mythos 5.2 exclusively for vetted organizations. Tight controls remain essential following an April incident where an unauthorized group accessed Mythos via a contractor.
What happens next
Enterprise customers will gain access to Claude Mythos through select rollouts, though no specific date has been announced yet. In addition to the security integration, Cisco also updated its Splunk tools designed to monitor AI agent expenses and performance quality. A new pricing tool named Tokenomics will enter private preview in December, affecting businesses tracking operational agent costs.
Frequently asked questions
Claude Mythos is an Anthropic model designed to plan attacks and generate testing ideas against AI apps and agents to find vulnerabilities.
Cisco announced the addition of Claude Mythos to AI Defense at the Cisco Live event in Melbourne.
No, customers do not have open prompt access to the model. It works only on a fixed job against systems the customer owns and has registered.
Every test is tied to a named person, recorded for audits, and can be stopped at any time. The model cannot scan the wider internet.
Palo Alto Networks launched always-on testing with Mythos in September, and IBM plans to add it to Concert Protect in early 2027.
Based on reporting by The Next Web. Read the original article
