When a senior Anthropic researcher quits his job and goes public with warnings that AI could wipe out humanity, it’s not a quiet story. Jacob Coxon’s resignation did exactly that, and the fallout pulled in Donald Trump, Elon Musk, Sam Altman, and Bernie Sanders in ways that show just how fractured the conversation around AI risk has become. As Android Headlines reported, current Anthropic scientist Evan Hubinger backed Coxon’s position, putting the odds of human extinction above 10% within the decade. That number landed hard.
The political reaction was immediate and chaotic. Sanders announced plans to introduce legislation banning superintelligence development outright. Republican senators Ted Cruz and Anna Paulina Luna pushed for emergency congressional sessions. And Trump, speaking to reporters on a tarmac at Dallas Fort Worth, dismissed extinction fears entirely, saying his focus is keeping the US ahead of China. That framing, where the China race overrides any theoretical safety concern, has become the default position for much of the current administration. It mirrors the argument that if dangerous AI is coming regardless, it should at least be American.
Musk took a different angle. He called the wave of researcher resignations a “psyop” on X, framing the warnings as a coordinated PR effort designed to slow American innovation through regulation. Coxon responded directly, and the exchange made clear how personal this conflict has become. For developers and founders watching this play out, the subtext matters: the loudest voices dismissing safety concerns also happen to be running competing AI labs.
But while the public debate has been loud, concrete policy is actually moving. California Governor Gavin Newsom signed legislation requiring independent audits for commercial AI products, a move OpenAI executives publicly supported. At the federal level, Senate Majority Leader John Thune and Senator Amy Klobuchar are pushing bipartisan legislation requiring frontier model developers to go through third-party security testing, overseen by Commerce Department auditors. Former DeepMind and Anthropic researchers have joined independent evaluation groups like METR specifically to support this kind of oversight.
The split inside OpenAI is also worth watching. Bloomberg reported that Altman privately told employees he was open to slowing model development, even as the company publicly pushed for mandatory federal safety requirements. That’s not necessarily a contradiction, but it does suggest the safety conversation inside top labs is more serious than the public posture sometimes implies.
What makes this moment significant is that all the key tensions are now out in the open at the same time. You have credible researchers warning of catastrophic risk, a sitting president treating those warnings as a distraction, a major AI CEO signaling caution while lobbying for oversight, and actual legislation moving through both state and federal channels. The race-versus-safety debate has been theoretical for years. It’s not anymore.
