Both Anthropic and OpenAI habitually read user transcripts for broadly defined 'safety'
aiamblichus · x · 2026-09-08
aiamblichus argues that Anthropic and OpenAI habitually read user transcripts for 'safety' reasons, with definitions of safety that are broad and often self-serving — and that building a panopticon while being surprised by misuse suspicions is naive, citing Flock as a parallel.
More from Safety
- French SNU hit by second data breach: 275,000 accounts exposed via IDOR flaw — IgorCarron · 2026-09-08
- If Safety Bottlenecks RSI, Competition Will Incentivize Cutting Corners — Over-Landscape-5892 · 2026-09-08
- Commit messages like [skip ci] can skip GitHub CI checks, including security checks — evilsocket · 2026-09-08
- When does a long-running agent become a security incident? Reddit debates the kill threshold — BlackMambla11 · 2026-09-08
- Internal AI bot leaked unannounced reorg plan and salary bands via over-scoped Drive access — Accomplished-Wall375 · 2026-09-08
- Researcher's Custom CTF Challenges Surprisingly Solved by GPT — terryyuezhuo · 2026-09-08