Anthropic's advanced AI model, Claude Mythos, is reshaping cybersecurity by enabling faster vulnerability discovery and resolution. With 150 organizations now using the model, it's becoming a critical tool in the fight against cyber threats.
TL;DR
- Anthropic's Claude Mythos AI model is transforming cybersecurity by enabling faster vulnerability discovery and resolution.
- The model is now used by 150 organizations, including Cisco, which has scanned 1.8 billion lines of code using Mythos.
What happened
Anthropic's advanced AI model, Claude Mythos, was released to an additional 150 organizations in June, bringing the total number of partners to around 150. The model is optimized for cybersecurity and healthcare applications and was originally released in April to a small group of vetted technology partners.
Anthropic established Project Glasswing, a consortium that gives limited, controlled access to Mythos to infrastructure providers, open-source developers, and major technology companies. The scheme was designed to enable defenders to find and resolve vulnerabilities faster than they could be identified by attackers.
Cisco, one of Anthropic's Project Glasswing partners, open-sourced its Foundry Security Spec, a model-agnostic "harness" for security testing, so that other vendors and enterprise security defenders could build similar workflows without starting from scratch.
Why it matters
Claude Mythos is the most prominent example of frontier AI models that can automate zero-day discovery at a scale and speed far beyond the capability of human teams. This makes it a critical tool in the fight against cyber threats.
The model is transforming cybersecurity by enabling faster vulnerability discovery and resolution. This is particularly important as attackers are using AI to accelerate the path from vulnerability discovery to exploitation.
However, critics note that while Claude Mythos is advanced, marketing claims that it is reliably breaking production systems overstate its capabilities. Security professionals are also complaining about the overly sensitive guardrails in Claude Fable that downgrade to Opus 4.8 upon requests to summarize a security-related blog post or even spell the word "exploit".
Key facts
- Anthropic's Claude Mythos AI model is now used by 150 organizations.
- Cisco has scanned 1.8 billion lines of code using Mythos.
- Anthropic established Project Glasswing to give limited, controlled access to Mythos to infrastructure providers, open-source developers, and major technology companies.
- Claude Mythos is optimized for cybersecurity and healthcare applications.
- The model was originally released in April to a small group of vetted technology partners.
- Anthropic is also offering Claude Fable 5, which is based on the same underlying technology but comes with strict guardrails that limit operations in "risky" cybersecurity domains.
- The US government applied export controls to Claude Fable 5 and Claude Mythos 5 on June 15, but the restrictions were lifted on June 30.
- Other AI vendors, such as OpenAI and 360 Security Technology, also have models aimed towards high-capability, security-oriented operations.
Context
Anthropic is a company that develops large language models for a variety of applications, including cybersecurity and healthcare. The company was founded in 2021 and is based in San Francisco, California.
The cybersecurity industry is facing a growing threat from AI-powered attacks. As AI becomes more sophisticated, it is becoming easier for attackers to discover and exploit vulnerabilities in software and systems.
The US government has been considering how to regulate the use of AI in cybersecurity. In June, the Trump administration applied export controls to Claude Fable 5 and Claude Mythos 5, but the restrictions were lifted on June 30.
