Anthropic's Mythos bug-finding model is being built into Cisco AI Defense and IBM Concert
Anthropic has begun letting security vendors embed its Mythos vulnerability-discovery model in their own products without giving customers the model itself. Cisco announced that the red-teaming tools in its AI Defense suite, which test AI models, agents and applications for weaknesses, will be able to call Mythos to probe customer setups. Cisco says the option comes under existing licenses at no extra cost. IBM said it will add Mythos to its IBM Concert tool, which scans applications for risks, to find weaknesses traditional scanners miss and to generate fixes for developer approval. The article also reports that Anthropic released Mythos 5.2 and Fable 5.2 the same day. This is a product and access-expansion story with no vulnerability, exploitation or victim involved, so the direct threat to defenders is low.
Detalhes técnicos
- Cisco's integration exposes Mythos functionality to end users through AI Defense's existing AI-powered red teaming, with Mythos as an additional testing option (per the article).
- IBM's planned integration combines Mythos-driven code scanning with Concert Protect's exposure management, including its proof-of-exploitability capability (ibm.com, announcement dated 9 November 2026).
- The article reports pricing of $8 per million input tokens and $40 per million output tokens for Mythos 5.2 and Fable 5.2, with a fast mode at double the price. These figures come from the article only; no vendor page was opened to confirm them.
Relatos divergentes
- The article says IBM "will also integrate" Mythos into Concert. IBM's own announcement (ibm.com, 9 November 2026) says IBM "plans to" bring Claude Mythos into Concert Protect, so the IBM integration is not yet confirmed as available. The IBM source is the more authoritative one on timing.
Linha do tempo
-
Anthropic's Mythos bug-finding model is being built into Cisco AI Defense and IBM Concert
ler a análise desta fonte ocultar
Anthropic has begun letting security vendors embed its Mythos vulnerability-discovery model in their own products without giving customers the model itself. Cisco announced that the red-teaming tools in its AI Defense suite, which test AI models, agents and applications for weaknesses, will be able to call Mythos to probe customer setups. Cisco says the option comes under existing licenses at no extra cost. IBM said it will add Mythos to its IBM Concert tool, which scans applications for risks, to find weaknesses traditional scanners miss and to generate fixes for developer approval. The article also reports that Anthropic released Mythos 5.2 and Fable 5.2 the same day. This is a product and access-expansion story with no vulnerability, exploitation or victim involved, so the direct threat to defenders is low.
Technical Details
- Cisco's integration exposes Mythos functionality to end users through AI Defense's existing AI-powered red teaming, with Mythos as an additional testing option (per the article).
- IBM's planned integration combines Mythos-driven code scanning with Concert Protect's exposure management, including its proof-of-exploitability capability (ibm.com, announcement dated 9 November 2026).
- The article reports pricing of $8 per million input tokens and $40 per million output tokens for Mythos 5.2 and Fable 5.2, with a fast mode at double the price. These figures come from the article only; no vendor page was opened to confirm them.
Conflicting Reports
- The article says IBM "will also integrate" Mythos into Concert. IBM's own announcement (ibm.com, 9 November 2026) says IBM "plans to" bring Claude Mythos into Concert Protect, so the IBM integration is not yet confirmed as available. The IBM source is the more authoritative one on timing.