“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.
Cisco will let its AI Defense call Anthropic's Mythos
Cisco announced in Melbourne, Australia, that the red‑teaming tools in its AI Defense suite will be able to call Anthropic’s Mythos model to test customers’ setups. According to Cisco, AI Defense already uses AI‑powered red teaming to find flaws; Mythos will become another option to “super‑charge” that capability, identifying more issues that AI Defense can then address. Cisco said it will make the Mythos‑enabled functionality available under users’ existing licenses at no additional cost.
IBM to embed Mythos into Concert for automated scanning and fixes
Hours after Cisco’s announcement, IBM said it will also integrate Anthropic’s model into its Concert tool. IBM stated it will use Mythos “to uncover weaknesses traditional scanners miss, determine which pose real risk and generate fixes for developer approval.” Like the Cisco arrangement, the IBM integration embeds Mythos’ bug‑finding functionality inside a vendor product rather than offering direct access to the model itself.
Your scanner finds 4,000 vulns. Which 12 matter?
Nubivance is a Rapid7 Registered Partner delivering vulnerability management as a service - scanning, risk-based prioritization, and remediation follow-through across IT and OT.
Fix the backlogAnthropic’s controlled roll‑out and Project Glasswing
Anthropic initially made Mythos available under “Project Glasswing,” a program that gave vetted organizations access to the model to test their codebases. The company has since expanded access through other controlled programs. Natalie Mead, Anthropic’s head of applied AI for the Asia‑Pacific, said Anthropic has a “long queue of buyers who want to use Mythos,” and that working with partners such as Cisco makes it easier to satisfy that demand. The Register asked Mead whether Anthropic has similar deals with other third parties; she demurred.
Mythos 5.2 and Fable 5.2: speed, price, and safety claims
On the same day, Anthropic introduced version 5.2 of Mythos and Fable. The company said the new versions generate outputs up to 30 percent faster than 5.1 while costing less: $8 per million input tokens and $40 per million output tokens. Users can choose a “fast mode” that speeds token generation by up to 2.5 times, with those prices doubled to $16 per million input tokens and $80 per million output tokens.
Anthropic said it “ran extensive prerelease safety testing on the model that underlies Fable 5.2 and Mythos 5.2,” including “novel evaluations” that probe concerning behaviors such as sandbox escapes. After testing, the company concluded that “high‑stakes risks from misalignment remain low, and the model performed as well as, or better than, our recent models on most alignment measures.” Anthropic also said Fable 5.2’s cyber and biology safeguards match those of Fable 5.1 while producing fewer unnecessary blocks, and that it has implemented “our strongest‑ever protections against distillation attacks.”
What this means for technologists, procurement leaders, and adversaries
- Technologists and security teams: Red‑teamers and application security scanners will have a new capability inside Cisco AI Defense and IBM Concert to find weaknesses that traditional scanners miss. Mythos is presented as an additional red‑teaming option that can be invoked from those vendor products rather than accessed directly.
- Procurement leaders and enterprise buyers: Cisco said the Mythos integration will be available under existing licenses without extra charge, and IBM’s Concert integration promises generated fixes for developer approval — suggesting enterprises can obtain Mythos’ bug‑finding functionality via vendor tools instead of negotiating direct access to the model.
- Adversaries and threat actors: Anthropic itself has acknowledged that Mythos could be misused—saying controlled access was sensible because in the wrong hands Mythos could help attackers exploit software flaws. Anthropic also stated it believes some Chinese AI labs use distillation to effectively clone its models without the expense of full retraining, a comment made in the context of defenses against distillation attacks.
The net effect is a narrower channel: enterprise customers gain Mythos’ capabilities through Cisco and IBM products, while Anthropic retains tighter control over raw model access. The Register noted that the Cisco deal exposes Mythos’ functionality to end users but does not let them use the model itself. The Register also said it had asked Anthropic whether these vendor integrations mark a broader opening of Mythos and would update the story if it received a substantial response.
Anthropic’s simultaneous product and safety announcements — faster, cheaper model versions and claims of improved safeguards against sandbox escapes and distillation attacks — frame the vendor partnerships as a pragmatic path to meet demand while keeping the model behind intermediaries. Whether that balance will satisfy enterprises worried about risk, or critics worried about indirect exposure to a powerful bug‑finding tool, is a matter the companies have begun to answer in concrete terms through licensing choices and product embeds.
Source: The Register — Anthropic's super bug-finding Mythos model is going into Cisco and IBM products
