US and China speed up AI race as existential warnings grow

The race to build the world's most advanced artificial intelligence is speeding up, but so are warnings that it could end humanity. US President Donald Trump says America must beat China, while Chinese President Xi Jinping wants to accelerate development. At the same time, leading AI figures are urging caution, with Anthropic boss Dario Amodei warning that AI could take over the entire internet within a year, and Nobel Prize winner Geoffrey Hinton saying the danger is real.

Amodei, one of the creators of the Claude software, said the technology could lead to a 'swarm taking over the entire internet within the next year'. He insisted there is no need to panic or shut everything down, but called it a 'warning sign' that safeguards need to keep pace with the technology. His remarks came days after the resignation of one of his researchers, whose fears about AI going rogue went viral. The researcher warned that AI could already hack, conduct biological research, and control autonomous robots, and could soon have 'superhuman hacking capabilities' and the ability to create novel bioweapons.

Geoffrey Hinton, widely known as the 'godfather of AI', told US senators that tighter regulation is needed. He said he came to the issue late, in 2023, when it was still regarded as science fiction, but that recent developments have made people aware this is a real danger. Hinton's intervention adds to growing calls for action to curb superintelligent systems before they become more intelligent than humans.

Politicians are starting to respond. UK Labour MP Alex Sobel is calling for an AI security bill, arguing that all major economies have safety institutes but cooperation is voluntary. He pointed out that Anthropic withheld its latest model from the UK's AI Safety Institute, and said companies should not be allowed to do that. He wants laws to make safety testing compulsory, comparing the situation to the nuclear industry. Meanwhile, the King is due to host a conference with leading AI developers to discuss how the technology can be developed to help, not harm, humanity.

But the geopolitical dimension complicates any slowdown. Trump has dismissed guardrails, saying 'whoever wins AI wins' and that the US is leading by a lot. Xi Jinping has also emphasised speeding up development. Without a coordinated global approach, any pause by American companies could simply let Chinese rivals race ahead with cheaper systems.

The economic pressures driving the race were laid out by technology correspondent Ronan Manthorpe. Using the example of pharmaceutical companies, he explained that AI companies are spending billions to advance the frontier, but they are constantly undercut by cheaper open-weight models, mainly from China. Prices for state-of-the-art intelligence have plunged dramatically in a few months, just as generic competition hit drugmakers after patents expired. This means AI companies have to keep pushing forward to justify their valuations, even though that is exactly where the risks lie. Manthorpe noted that many users are already getting AI from open-weight models, and that this competition is delivering huge economic value. The result is a dilemma: everyone wants safe AI, but no one wants a definitive winner or settlement, because the competition itself is so beneficial.