Anthropic has unveiled Claude Mythos, a new AI model that claims to spot cybersecurity bugs faster than human experts. In an early‑April blog post, the company warned that if such power fell into malicious hands, the fallout could cripple economies, public safety and national security. The announcement came while the model remains locked behind a selective release.

Critics argue Anthropic’s scare tactics distract from the company’s broader impact on labor, energy use and data privacy. Industry voices like Sam Altman and Dario Amodei have warned that fear‑mongering boosts stock prices while sidelining regulation. Still, the company cites partnerships with over 40 firms to patch vulnerabilities before attackers exploit them, framing Mythos as a defensive tool.

Security researchers question Mythos’ superiority, noting the absence of false‑positive metrics and comparisons to established static‑analysis tools. Heidy Khlaaf, an AI‑safety expert, cautions that without transparent benchmarks, claims of thousands of high‑severity bugs remain unverified. Even if Mythos can scan code faster, the real danger lies in the potential for adoption without rigorous testing, which could expose flaws.