Artificial Intelligence
Anthropic moves Claude Mythos 5 into cyber defence
Anthropic said Claude Security was running on Claude Mythos 5 in public beta for Enterprise customers, while partner security products used the model to support hospitals, utilities and banks.
- Anthropic said Claude Security was running on Claude Mythos 5.
- It said the scanner checked codebases for vulnerabilities, assigned CWE classifications and severity ratings, and suggested patches.
- Anthropic said partner security products used Mythos 5 for hospitals, utilities and banks, and end users did not interact with the model directly.
- Anthropic said some partners already used Claude Opus for security tools and were expected to switch to Mythos 5.
- Anthropic said Claude Mythos was its most capable model, especially for cyber tasks, and that the rollout was meant to help defenders without giving attackers new AI options.
Anthropic said Claude Security was running on Claude Mythos 5. The scanner looked for vulnerabilities in codebases, suggested fixes, and was available in a public beta for Enterprise customers.
Each result included a CWE category, a severity rating and a proposed patch, and scans were counted as normal token usage. Anthropic said humans still had to approve every patch.
The company also said it had added Mythos 5 to partner security products used for hospitals, utilities and banks. It said end users would only see outputs such as suggested patches, not the model itself.
Anthropic said several partners already used Claude Opus for security tools and were expected to move to Mythos 5. It also said security vendors could apply for partnership access.
Anthropic described Claude Mythos as its most capable model, especially for cyber work. It said the rollout was intended to support defenders without giving attackers new AI-enabled capabilities.
Named in this story
Companies
- Anthropic
- said it was running Claude Security on Claude Mythos 5 and was embedding the model in partner products
Products and systems
- Claude Security
- Anthropic’s security scanner
- Claude Mythos 5
- the model used by the scanner and by partner security products
- Claude Opus
- the earlier model several partners were already using for security tools
How the source tells it
The piece had a promotional register, built around Anthropic’s own superlatives and a reassurance that the rollout would help defenders without opening new options for attackers.
- vendor boosterism repeated self-applied superlatives without outside validation
- reassurance a direct safety assurance that the deployment would aid defenders rather than attackers