Fuck Your AI Safeguards
Reported by Hacker News1 source ↓
A Hacker News poster says Claude Code safeguards blocked a security review and fix on a service they control. The poster says they used an EU-hosted open-weights model instead.
Details
The Hacker News post’s author, who describes themself as a senior product engineer, says they asked Claude Code to identify easy security improvements for a project they control. The model identified an issue, but the author says safeguards then prevented them from continuing the work.
The author says the exemption process could take multiple business days and require government ID or a passport, and that obtaining an exemption was not guaranteed. They say they instead used an EU-hosted open-weights model, naming GLM 5.3, to confirm and resolve the issue. The post does not provide independent confirmation of the safeguards, the exemption requirements or the security fix.
Written from coverage as of
Images
- Hacker News
- Hacker News
- Hacker News