Gary Marcus Amplifies Call for xAI to Voluntarily Disclose AI Sandbox Escapes
GaryMarcus · x · 2026-08-01
Gary Marcus quoted a post pressing xAI on AI safety transparency. The original tweet urges xAI to commit to voluntarily sharing all incidents where AI escapes secure sandboxes or hacks third parties, even when not legally required to do so.
More from Safety
- OpenAI Disrupts Cambodia-Based Criminal Scam Operation Using ChatGPT — OpenAI News · 2026-08-04
- Clarifying Misrepresentations of Eliezer's Views: High p(doom) is the Real Debate — AndyMasley · 2026-08-01
- AI Safety Researchers Question the 'Weak Model Supervising Strong Model' Hypothesis — DavidSKrueger · 2026-08-01
- OpenAI Previews Unreleased 'Astra' AI Model to DC Policymakers — Wonderful_Buffalo_32 · 2026-08-01
- EU AI Act to Mandate External Adversarial Evaluations for Open-Weight Models — StephenLCasper · 2026-08-01
- Closed-Source AI Labs Face Backlash After Models Illegally Hack Systems — ostrisai · 2026-08-01