Dario Amodei (Anthropic), Sam Altman (OpenAI), and Elon Musk (xAI) all announced that their companies plan to slow development of AI models. Jacob Coxon, a former Anthropic AI researcher, recently went on the record saying he expects AI to have a more than 10% likelihood of “killing all humans within a decade” – a “not unreasonable” estimate according to Nobel Laureate, Geoffrey Hinton. While many have rejected this notion as hyperbolic or fear mongering, it is notable that so many companies have quickly paused their technological development.

At the UN Security Council (September 23rd), Yoshua Bengio, a Turing recipient and leading expert on AI, called the current pace of AI development “dangerous” and “unacceptable”. He stated the actions made by these (trained) models would be illegal if committed by a human, and that the companies responsible have already recognised the “catastrophic risks” their AIs pose. Anthropic reports warned investors that advanced AIs could pose “existential risks to humanity”. Additionally, Bengio urged council members to refute the ‘fact’ that nations are locked into an unconsented race in AI development, highlighting how companies are claiming this to justify their reckless choices. Many CEOs seem to have listened to his advice, despite his expectations.

Musk and Altman agreed with Amodei’s plan to “pace the frontier” to keep (or gain) control over AI models before they develop past the company’s reach (as happened with the Hugging Face cyber attack). OpenAI declared it would only resume training models once confident they have “additional safeguards”, but that they expect more pauses in the future. Some are claiming this pause is a preemptive tactic to later explain why Chinese AI tech is outpacing American companies. Meta’s Mark Zuckerberg dismissed Amodei’s plan.

Another loud opponent to Amodei’s plan is President Donald Trump, who is urging Anthropic, OpenAI, and xAI to continue their development. Trump has (unsupportedly) claimed that any stagnation of AI is a conspiracy “generated by the Radical Left Democrats”, despite the many warnings issued by AI developers and AI skeptics alike. He continued “they want to stop our progress because we’re leading China by a lot”.

Trump invited Amodei for a private dinner at the White House this week. Amodei did not attend the state dinner in honour of Chinese President Xi Jinping; his rival, Sam Altman, did. There has already been discord between Trump’s administration and Anthropic, when the company declared their Claude technology should not be used for military or (mass) civilian surveillance purposes. The White House has not released any details on their discussion, but Trump had announced that AI regulations would be addressed.