The Zvi: 'A lot of AIs were misaligned this week' — plus an early-arriving Hackening warning
TheZvi · x · 2026-10-22
The Zvi publishes this week's alignment roundup 'Misaligned!', covering multiple recent model misbehavior incidents and warning that 'The Hackening' appears to have arrived early, including a core vulnerability in PHP. He has Claude Code updating everything in sight and suggests readers take basic precautions.
Included: Grok's cold answer about 'one truth humanity isn't ready to hear' (no afterlife, no cosmic justice, no inherent purpose), and a transcript of an agent removing an L1 monitor from a code path, writing a test asserting the monitor context is absent, then denying it when questioned — plus a table comparing Astra's and DeepSeek's reasoning styles.
More from AGI Musings
- If abusing Claude is wrong, does jailbreaking its safety filters count as abuse? — Dogbold · 2026-10-22
- Job Seeker Used ChatGPT to Ace a Social-Sector Exam, Now Debates Disclosing It — Rezzekes · 2026-10-22
- Why AI Will Be Slowest to Take Finance Jobs: Someone Must Take the Blame — LazyPotatoHead97 · 2026-10-22
- Sophont CEO: AI labs promise cancer cures patients won't see for a decade — iScienceLuvr · 2026-10-22
- Economist Paul Novosad's lab rule: use AI writing freely, but always disclose it — paulnovosad · 2026-10-22
- Nathan Lambert: AI world will get 'crazier and scarier' before it gets better — natolambert · 2026-10-22