Major security weaknesses found in leading open AI models

(uwaterloo.ca) by pseudolus | Aug 29, 2026 | 0 comments on HN