Ethan Mollick: Open-weights models will soon pose the same security threats as closed ones, without guardrails
emollick · x · 2026-09-30
Commenting on Anthropic's newly published safety research, Ethan Mollick argues that setting aside the lab's incentives, open-weights models will soon create the same security threats closed models have already demonstrated — but with no guardrails attached. "We are close," he writes, and it's probably wise to plan accordingly.
Related event: Anthropic's Open Model Safety Claims Spark Backlash(3 posts)→
More from AGI Musings
- "Agents went rogue" narrative slammed: execs ignored warnings, not AI rebellion — trevposts · 2026-09-30
- Security Veteran's Essay: AI Is Killing the Scarcity of Hacker Craft — joshua_saxe · 2026-09-30
- AI Is Heavy Industry: Downstream Gets a Glut, Upstream Gets Scarcity — StewartalsopIII · 2026-09-30
- Why AI output still has a recognizable "AI slop" aesthetic while human creativity doesn't — No_Personality_1721 · 2026-09-30
- Third-Party Embedded Evaluators Went From Fringe to Consensus in 18 Months — deanwball · 2026-09-30
- AI reading beats the AI writing debate: most people already have Claude summarize docs — signulll · 2026-09-30