Anthropic's Claude Mythos AI model heightens fears that hackers without technical skills can use AI to find and exploit software vulnerabilities

Anthropic released Claude Mythos, an AI model described as finding vulnerabilities in software it is pointed at. A week after announcing Mythos, the company released Claude Opus 4.7, which built in safeguards meant to block malicious cybersecurity requests. Security professionals can apply to Anthropic's Cyber Verification Program to use the model defensively.

AI systems are dramatically lowering the barrier to finding bugs. A security researcher who previously took weeks or months to find a high-impact vulnerability in new code now takes hours using AI bug-finding tools. This makes it feasible for attackers to find zero days (previously unknown vulnerabilities) in uncommon software that would have required too much effort to exploit manually.

Security experts warn that this capability extends hacking power to 'script kiddies'—people without deep technical knowledge who previously relied on pre-written exploits from the internet. Attackers can now point AI at unknown software during an active intrusion and have it find a vulnerability with minimal human guidance, operating at 'machine speed' rather than at human pace.