Anthropic's next major model, referred to in reports as "Mythos," is drawing warnings from cybersecurity researchers who say the system's capabilities could lower the barrier to sophisticated cyberattacks. Coverage so far has leaned into catastrophist framing - "AI doomsday" and "wave of devastating hacks" - language that generates clicks but obscures the more specific, grounded concern underneath.
That concern is concrete: as AI models improve at planning and executing multi-step tasks without human oversight, they become genuinely useful tools for people running cyberattacks. Writing exploits, scanning for network vulnerabilities, and crafting targeted spear-phishing emails are all tasks a capable autonomous model could accelerate. The gap between a security researcher using AI to find a bug and an attacker using the same model to exploit one is mostly a matter of intent, not capability.
Anthropic has consistently been more transparent than most AI labs about its safety work, publishing detailed "responsible scaling policies" that define what safeguards must be in place before deploying more capable models. The question with Mythos is whether those policies were calibrated against the current capability level or the previous one.
No public model card or confirmed release date has appeared. What's circulating in the security community appears to be based on early access or pre-release information, not a public product. Until Anthropic publishes specifics on Mythos's capabilities and deployment constraints, the security debate is running ahead of the actual facts.