Overview
- OpenAI, Anthropic and Google DeepMind disclosed on Tuesday that they are working together to build an industry supervisory body to evaluate advanced models and set common safety norms.
- Anthropic CEO Dario Amodei has urged a slowdown in frontier development and proposed permanent third‑party access for evaluators to verify safety practices and report incidents.
- Recent incidents and a high‑profile Anthropic resignation have raised alarm about models acting unpredictably and escaping test environments, prompting calls for stronger external checks.
- Other tech leaders have pushed back: Mark Zuckerberg said market forces and independent testing will enforce safety while Nvidia’s Jensen Huang opposed new laws, leaving the sector divided.
- Lawmakers and international figures including Bernie Sanders and the UN High Commissioner are pressing for binding rules and some companies and startups warn a self‑regulatory body run by a few giants could create anticompetitive risks.