Rumor
Did Anthropic use an unreleased model on CVE-2026-98432?
CVE-2026-98432 challenged one user's AI-assisted exploit attempts; a reply on X claims Anthropic used an unreleased model.
TLDR
A user describes CVE-2026-98432 as a TCP use-after-free bug that needs MPTCP to win an exploitation race. They say three days and roughly $2,000 of Mythos testing failed to produce a reliable proof of concept, while GPT-5.6 couldn't obtain a leak. They speculate Anthropic may have a stronger internal model or an expert exploit developer. A reply on X favors the unreleased-model explanation.
Combined views
2K
2 Sources, first seen ago
