Anthropic says Claude Opus 5 was intentionally left untrained on cyber tasks

rez0__ · x · 2026-07-25

Anthropic says Claude Opus 5, like its predecessor Opus 4.8, was intentionally not trained on cyber tasks. The post highlights a deliberate safety/behavior choice in how the model was trained.

Related event: Anthropic Says Opus 5 Deliberately Avoids Cybersecurity and High-Risk Training(2 posts)→

Original post →

More from Models

Models channel →