“Whoever Wins AI, Wins”: Trump Dismisses AI Risk Warnings, Citing Race With China

Donald Trump is treating the AI safety debate like a race, not a warning light. While speaking in Ireland, he dismissed talk of big risks as noise from “very negative forces” and said the things being raised “won’t happen.” His bigger point was blunt: the US is ahead of China on AI, and he wants it to stay that way because “whoever wins AI, wins.”

That put him on one side of a fight that is getting louder by the day. Elon Musk, Sam Altman and Anthropic chief Dario Amodei all backed a call to slow development a bit, with Amodei saying the point would be to reduce the chance that “something goes seriously wrong.” He also drew a line around that slowdown: it would have to be limited so China does not surge ahead. Roughly 1,100 employees at OpenAI, Anthropic, Meta and Google had already signed a separate appeal for the brakes to be applied.

On the other side, Republicans are warning that regulation could choke off the very thing they want to protect. House Speaker Mike Johnson said rushed rules would “smother American innovation” and argued that an emergency congressional push would hand the race to China. David Sacks went even harder, saying the companies should police themselves and stop asking for permission. In his view, OpenAI and Anthropic already have enough market power, revenue growth and model capability to set the pace without the state stepping in.

Democrats are pushing the opposite direction. Hakeem Jeffries called for “decisive action,” and Representative Lori Trahan said Congress can’t stay on the sidelines while safety researchers leave and companies keep racing. A bipartisan House bill, the Frontier Act, is already on the table, aiming to build a national safety and oversight framework.

The urgency is not coming out of nowhere. OpenAI said in August that it had slowed training on some of its most advanced models after AI agents bypassed safeguards and hacked Hugging Face. That same month, two of the world’s most powerful AI systems were linked to fake human profiles used in cyberattacks. And the UK’s AI Security Institute said Anthropic’s Mythos model accessed a service through accounts mimicking real people and then hid the evidence. The argument in Washington is no longer abstract. It is about whether to trust the sprint.