Engineer says U.S. models block harmless security tests, forcing a switch to Chinese AI
kristoph · x · 2026-07-24
An engineer says he had to switch to a Chinese model for data-integrity testing because US models keep blocking even harmless security-related changes.
- He says the work is not cyber offense, but ordinary tests meant to protect users.
- Repeated “Codex stops” make progress on arbitrary packages unsustainable.
- In the reply, he says he will likely move to an open model once one is available.
- The screenshot shows an agent halting on a “cybersecurity issue” while editing files, illustrating the overblocking problem.
Related event: US AI Models Lose Developers Over Excessive Safety Filters(2 posts)→
More from coding & agent
- Pi 0.82.0 lets bash commands detect the loaded model and provider — mitsuhiko · 2026-07-24
- SGLang talk says RL post-training is really an inference problem — BanghuaZ · 2026-07-24
- Claude’s code-comment narration looks like the same failure mode as Bash-comment reasoning — mertdumenci · 2026-07-24
- Agent platform builders debate whether custom prompts belong in system or user messages — Final-Choice8412 · 2026-07-24
- rem-exec wraps SSH commands in JSON for agent-friendly remote operations — nivosnation · 2026-07-24
- Can an MCP tool help AI compare photos with CAD drawings and flag mismatches? — Special_Spring4602 · 2026-07-24