Anthropic's Fable 5.2 blocks fewer security tasks as Mythos 5.2 finds more bugs
Security researchers with access to Mythos 5.2 gain a stronger tool, while general users of Fable 5.2 see fewer false refusals on security work.
Anthropic has released Claude Mythos 5.2 to selected partners and Claude Fable 5.2 to general users, with gains in coding and reasoning as well as cybersecurity changes. Mythos 5.2 is positioned as the stronger tool for security research, outperforming Fable 5.2 at locating vulnerabilities, reproducing them and drafting fixes. In Anthropic's CVE Rediscovery evaluation, Mythos 5.2 flagged roughly a tenth more vulnerabilities than Claude Opus 5.5 and about 14 percent more than Mythos 5.1. Fable 5.2 carries additional cybersecurity and biology safeguards that block high-risk work such as penetration testing and exploit generation, routing those requests to Opus 4.8 instead. Anthropic says Fable 5.2 cut unnecessary cybersecurity blocks by about 60 percent versus Fable 5.1, though it concedes its evaluations cannot catch every failure before release.
Why it matters
The split leaves general users with a model that still refuses high-risk work such as penetration testing and exploit generation, routing those requests to Opus 4.8, and gives vetted partners the stronger model for vulnerability research. Anthropic's own evaluations are not claimed to catch every failure before release, so the balance between fewer blocks and safeguards will be judged in use.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.