Opus 5 becomes the hardest model to jailbreak, with near-zero attack success when combined with defenses
Opus 5 is a great model for coding, data analysis, design, biology, knowledge work. More than any o...
AI BriefingAnthropic’s Opus 5 achieves new SOTA results across evaluations in coding, data analysis, biology, and more. More importantly, it has become the company’s hardest model to prompt-inject to date. With multiple layers of defense—including strong alignment, injection detection, and Claude Code’s Auto Mode—the success rate of prompt-injection attacks has fallen to approximately 0%.
Source: Boris Cherny · x.com