The AI model Mythos was found to have a CVE already present in its training data, which raises concerns about its security implications.

Mythos 'Discovered' a CVE in Its Training Data and That's Still Worrying L: https://rival.security/posts/mythos-discovered-a-cve-already-in-its-training-data---and-thats-still-worrying C: https://news.ycombinator.com/item?id=48072671 posted on 2026.05.09 at 03:17:05 (c=0, p=4)

Mythos 'Discovered' a CVE in Its Training Data and That's Still Worrying https://rival.security/posts/mythos-discovered-a-cve-already-in-its-training-data---and-thats-still-worrying #security

Mythos Discovered a CVE in Its Training Data – and That's Still Worrying Anthropic의 Claude Mythos AI가 FreeBSD 네트워크 파일 시스템의 오래된 스택 버퍼 오버플로우 취약점(CVE-2026-4747)을 발견하고 원격 코드 실행에 성공했다는 주장이 나왔다. 그러나 이 취약점은 2007년에 MIT Kerberos에서 이미 발견되어 패치된 CVE-2007-3999와 매우 유사한 코드 기반에서 비롯된 것으로, Mythos가 완전히 새로운 취약점을 발견했다기보다는 학습 데이터에 포함된 오래된 버그를 재발견한 것으로... https://rival.security/posts/mythos-discovered-a-cve-already-in-its-training-data---and-thats-still-worrying

Mythos Discovered a CVE in Its Training Data – and That's Still Worrying L: https://rival.security/posts/mythos-discovered-a-cve-already-in-its-training-data---and-thats-still-worrying C: https://news.ycombinator.com/item?id=48091882 posted on 2026.05.11 at 03:03:18 (c=0, p=4)

Mythos 'Discovered' a CVE Already in Its Training Data - and That’s Still Worrying https://lobste.rs/s/zwp0dp #security #vibecoding https://rival.security/posts/mythos-discovered-a-cve-already-in-its-training-data---and-thats-still-worrying