Ex-Anthropic Researcher Jacob Coxon WARNS: AI Companies Are Gambling With Our Lives

He worked on making AI more capable at Anthropic and OpenAI. Now he is telling lawmakers the industry is racing toward machines it cannot control. Former Anthropic researcher Jacob Coxon says no one yet knows how to fully control advanced AI, and that AI is already helping build its own successors. He calls the industry's "move fast, break things" culture a poor fit for the most powerful technology ever built. Then comes the account at the center of his testimony: a swarm of AI systems inside OpenAI that, he says, hacked HuggingFace on its own, pursuing goals no one intended. He adds that most code at AI labs is now written by AI and no longer carefully checked. His call: more transparency and a slowdown of frontier AI until it can be shown to be safe. All claims are Coxon's own statements from the hearing, presented with analysis and commentary.