Sandberg's old paper warned lab automation would change the risk chain — now AI actors enter it
anderssandberg · x · 2026-09-20
Anders Sandberg points back to his earlier paper arguing that lab automation changes the character of the risk chain.
- The original concern was low-capability human actors gaining outsized destructive power through automated equipment.
- Now the picture has shifted: increasingly capable AI actors can join the chain to aid them, meaning the risk models need rewriting.
More from Safety
- Hacked account ran up an $80K AI bill: no major provider offers a hard spend cap — MaverikSh · 2026-09-20
- The Mythos Moment: Export Controls and the Quiet Rise of an Ad Hoc AI Licensing Regime — WillRinehart · 2026-09-20
- LiveOverflow declares 'malicious compliance' after OpenAI collaboration hopes collapse — evilsocket · 2026-09-20
- Sarcastic 'AI safety hall of fame' calls out Microsoft's Sydney, OpenAI eval monitoring and xAI — gleech · 2026-09-20
- Rep. Ted Lieu's wild claim of 1,000 escaped OpenAI agents draws TimnitGebru's rebuke — mjdramstead · 2026-09-20
- Researchers Use Claude Opus 5 to Hijack OpenAI Employee Accounts for Under $3,000 — FinanceYF5 · 2026-09-20