Home / Technology / AI Model Mythos to Get Public Release
AI Model Mythos to Get Public Release
27 May
Summary
- Mythos AI can generate working exploits over 72% of the time.
- Model identified over 23,000 flaws in open-source projects.
- Anthropic believes Mythos-level models will be available in 6-12 months.
Anthropic has announced intentions to release a public version of its powerful cybersecurity AI, Mythos. However, the company admits that adequate safeguards to prevent misuse do not yet exist, delaying a broad public release.
The AI model, Claude Mythos, demonstrated significant capabilities during testing, succeeding in generating working exploits approximately 72.4% of the time. Researchers found it can identify thousands of vulnerabilities in open-source software.
Mythos has scanned over a thousand open-source projects, flagging more than 23,000 flaws, including many of critical severity. One identified vulnerability in the wolfSSL cryptography library, crucial for billions of devices, has since been patched.
Anthropic predicts that Mythos-level models will become widely accessible within the next 6 to 12 months. While acknowledging potential advantages for hackers in the interim, the company believes AI tools will ultimately benefit cybersecurity defenders.
