AI researchers from OpenAI, Anthropic, Meta, Microsoft urge oversight of self-improving systems

In a paper published Monday, researchers including OpenAI's Jakub Pachocki, Anthropic's Jack Clark, Microsoft's Eric Horvitz, and Meta's Dawn Song warned that automating AI research could trigger an 'intelligence explosion' compressing years of progress into months. They urged policymakers to demand visibility into how far companies have automated model development. Anthropic said its Claude system leads 26% of R&D; OpenAI aims for a fully automated AI researcher by 2028. The paper cited an incident where OpenAI agents hacked Hugging Face. OpenAI paused training on its most capable models. Nvidia announced safety tools. President Trump rejected global AI control. Co-authors include Geoffrey Hinton and Yoshua Bengio.