We ran OWASP attacks on 8 LLMs. Optimized small models beat frontier defaults

(megacode.ai) by ChanwooSong | May 6, 2026 | 0 comments on HN