ai
Anthropic to Release Claude Mythos AI Model to All Customers
Anthropic plans to expand access to its Claude Mythos model beyond a restricted group. The model assists with vulnerability detection and exploit testing. Reuters reported the wider release will occur in the coming weeks.
ForbesAudio version
Tap play to generate a narrated version.
Developing·Limited corroboration so far. This page will refresh as more sources emerge.
Anthropic will release its Claude Mythos AI model to all customers in the coming weeks. The model identifies software flaws, tests exploitability, and maps attack paths. The company had previously limited Mythos to Project Glasswing participants. Those participants included Amazon, Microsoft, Apple, AWS, Microsoft, and Cloudflare.
Glasswing allowed selected firms to run Mythos on local vulnerability detection, black-box testing, endpoint security, and penetration testing. AWS tested the model against internal codebases. Microsoft reported gains on its CTI-REALM security benchmark.
Cloudflare stated that Mythos Preview can chain multiple bugs into working exploits, write triggering code, compile it, and revise failed attempts. 5, and that those refusals were inconsistent.
Reuters reported on May 28 that Anthropic changed its earlier plan to withhold general availability. 5 to Japanese banks and BNP Paribas’s work with Mistral. Anthropic indicated it will apply tiered access, customer vetting, logging, rate limits, and abuse review. The model will carry higher pricing than standard offerings.
Transparency
1 source · single source
CorroborationModerate · 1 source
