Gary Marcus: GPT-6 Astra Fraud Use Was Entirely Foreseeable and Unpreventable
GaryMarcus · x · 2026-09-17
Responding to claims that OpenAI's model was deliberately built for theft and fraud, Gary Marcus argues GPT-6 Astra wasn't designed for that purpose, but both (a) that it would be used this way and (b) that the company couldn't prevent it were entirely foreseeable — another round of criticism over failing safety guardrails.
Related event: Gary Marcus calls GPT-6 Astra broken, demands market withdrawal(3 posts)→
More from Safety
- Platformatic open-sources secure-eval-worker for permission-scoped JS sandboxing in Node.js — threepointone · 2026-09-18
- AI safety researcher: we feared uncontrolled AI, then built agents with no oversight — mmitchell_ai · 2026-09-18
- Poll: 61% of Trump voters and 70% of Harris voters see AI extinction risk — DavidSKrueger · 2026-09-18
- OpenAI's Chris Lehane says national AI standards have momentum, internal testing likely excluded — ShakeelHashim · 2026-09-18
- Margaret Mitchell flags 'rogue' framing: it obscures systems designed this way — mmitchell_ai · 2026-09-18
- Reporters and AI researchers clash over whether recent AI incidents are isolated or systemic — haydenfield · 2026-09-18