Anthropic has launched two new versions of its Claude AI model: a full-scale cybersecurity-focused system called Mythos, and a safer, trimmed-down release named Fable 5 aimed at general users. The dual rollout marks a deliberate split between high-risk security applications and everyday use.
Two models, two purposes
The company says Mythos is built specifically for cybersecurity tasks. It's the complete version of the model, meaning it has the full range of capabilities needed to analyze threats, write defensive code, or simulate attacks. The other model, Fable 5, is a safer variant that limits certain functions to reduce the risk of misuse. Anthropic hasn't detailed exactly which capabilities are cut from Fable 5, but the distinction is clear: one model is for professionals who need maximum power, the other for the public who need guardrails.
Why the split matters
Releasing two flavors of the same core model lets Anthropic serve very different audiences without forcing everyone into one box. Security teams get the raw tool they need to keep up with fast-moving threats. Regular users get an assistant that won't accidentally generate dangerous code or help with hacking. The move mirrors a growing trend among AI labs to offer tiered access based on risk, especially as regulators push for safer deployment of powerful models.
What's in Mythos
Mythos is designed to handle the kind of work that requires deep understanding of system vulnerabilities and attack patterns. It can analyze logs, suggest patches, and even write exploit code—though the company likely has controls in place to prevent abuse. For now, Anthropic hasn't said whether Mythos is available only to vetted customers or to anyone willing to pay. The model's name, drawn from Greek mythology, hints at its role: a tool for confronting complex, hidden threats.
Fable 5 for everyday users
Fable 5, by contrast, is built for safer interactions. It retains the conversational and analytical strengths of Claude but with restrictions that prevent it from offering step-by-step guidance on illegal activities or generating malicious code. The name suggests a storybook quality—a model meant to inform, not to weaponize. General users who want an AI assistant for writing, research, or coding help can use Fable 5 without worrying about crossing ethical lines.
Both models are now available, but Anthropic hasn't disclosed pricing or the exact cutoff between what Mythos can do that Fable 5 cannot. That gap will likely be tested by security researchers and curious users in the coming weeks.
