Here’s How an AI Slowdown Could Actually Be Enforced

Researchers argue that slowing AI development remains an unsolved puzzle, citing the need for independent evaluations, tighter regulations, and new measurement methods. Anthropic’s latest report shows its Claude model now accounts for 26% of research output and 6% of its compute budget is dedicated to safety, while other labs propose third-party audits to detect misbehavior. Experts emphasize that effective control will require external funding and expertise beyond the AI companies themselves.