Anthropic plans to release its advanced cybersecurity AI model, Mythos, to U.S. allies and then the public within the next 6-12 months, pending the development of adequate safeguards.
TL;DR
- Anthropic's Mythos model, which excels at finding cybersecurity vulnerabilities, will first be expanded to U.S. and allied governments before a broader public release.
- The company acknowledges that sufficient safeguards against misuse do not yet exist, but believes Mythos-class models will be widely available within a year.
- Mythos has already identified over 23,000 flaws in open-source projects, including a critical vulnerability in the widely used wolfSSL cryptography library.
What happened
Anthropic has announced plans to eventually release a public version of Mythos, its AI model with advanced cybersecurity capabilities. The company first needs to develop adequate safeguards, which it admits do not yet exist. According to a report by The Register, Anthropic intends to expand access to U.S. and allied governments before pursuing a broader public release of 'Mythos-class models' in the 'near future.'
In an update to Project Glasswing, its limited-access security program, Anthropic stated that it believes 'Mythos-level models will become widely available in the next 6-12 months.' The company was candid about the timeline's uncertainty, noting that no company has yet built sufficient safeguards to prevent misuse of such models.
Why it matters
For developers and startups, the release of Mythos could provide a powerful tool for identifying cybersecurity vulnerabilities, potentially enhancing the security of their products. However, the lack of adequate safeguards raises concerns about the potential misuse of such advanced AI models.
Investors should note that while Mythos represents a significant advancement in AI-driven cybersecurity, the company's admission that safeguards are not yet in place may pose risks. The competitive advantage of Mythos could be substantial, but the timeline for its public release remains uncertain.
Key facts
- Mythos was initially revealed in April and succeeded at generating working exploits 72.4% of the time during testing, compared to just above 0% for Claude Opus 4.6.
- The model has scanned over a thousand open-source projects and identified more than 23,000 flaws, including over 6,000 rated high or critical severity.
- One of the identified vulnerabilities was in the wolfSSL cryptography library, used by billions of devices, which has since been patched.
- Anthropic plans to expand access to Mythos-class models to U.S. and allied governments before a broader public release.
- The company believes Mythos-level models will be widely available in the next 6-12 months, but acknowledges that sufficient safeguards do not yet exist.
- Open-source maintainers have reportedly asked Anthropic to slow its disclosure rate, citing an overwhelming number of bug reports.
- Anthropic suggests that AI tools will ultimately benefit cybersecurity defenders, though hackers may have a temporary advantage.
Context
The release of Mythos comes at a time when AI tools are becoming increasingly proficient at finding exploits, sparking a zero-day bug discovery crisis. Some bug bounty programs have shut down entirely due to the overwhelming number of vulnerabilities being identified.
Anthropic's approach to cybersecurity with AI models like Mythos highlights the dual-use nature of advanced AI technologies. While they can significantly enhance security, they also pose risks if not properly safeguarded.
The broader AI landscape is seeing a surge in AI-driven cybersecurity solutions, with companies racing to develop models that can identify and mitigate vulnerabilities more efficiently than traditional methods.
