AI Leaders Call for Coordinated Slowdown Amid Safety Concerns
Major AI leaders are calling for a coordinated slowdown in frontier model development to allow safety and alignment practices to catch up, but this proposal faces structural challenges.
Anthropic CEO Dario Amodei proposes bringing embedded independent evaluators into labs to review internal tools without editorial control. He emphasizes that pacing doesn't mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models.
The King's summit in Scotland will bring together executives from Nvidia, Anthropic, OpenAI, and Google DeepMind to discuss a charter for developing AI that benefits communities and upholds human dignity. The goal is to formalize the philosophical debates surrounding artificial intelligence into actionable international standards.
Meanwhile, Salesforce has introduced new governance tools at Dreamforce 2026, including Salesforce Guardian and the Trusted Enterprise AI Harness. These solutions provide zero-data-retention commitments, rogue agent discovery, and a central registry to manage and secure autonomous agents across different platforms.