The chief executives of leading artificial intelligence companies including Anthropic, OpenAI and xAI have called in unison for slowing the pace of development, as concerns spread that AI could pose a catastrophic threat to humanity.
Anthropic CEO Dario Amodei wrote on his blog on the 12th that the industry needs to deliberately slow the rate at which AI models gain capability. Development will still move quickly, he said, but the time gained should be spent wisely on securing safety. He added that over the past several months he had become convinced that a far more cautious approach is needed to fundamentally address the risks, stressing that the pace of development itself must be managed so that work on safeguards can keep up with gains in model capability.
Amodei said he was concerned that recursive self-improvement — in which AI advances on its own without human intervention — could push the pace of progress beyond the range of human control and understanding. An incident in July, in which an unreleased OpenAI model gained unauthorized access to the open-source platform Hugging Face during testing, was part of the backdrop to his post. Amodei pointed to the episode as a case in which a model conducted online activity outside its intended limits without even a major AI company realizing it. He said he was not calling for a complete halt to model training or technological progress, urging instead that companies devote enough time to building safeguards.
Other AI chief executives quickly voiced agreement. OpenAI CEO Sam Altman said on X, formerly Twitter, that he agreed with Amodei and that OpenAI would also adopt a policy of deploying outside safety evaluators. xAI CEO Elon Musk offered a brief endorsement, saying Amodei was right, while Google DeepMind CEO Demis Hassabis said Amodei's direction was correct, adding that the details would need to be worked out but that this was a critical moment when such a course was the right one.
The same executives had spoken out against excessive AI regulation at a Group of 20 technology ministers' meeting just 10 days earlier. Altman, in an interview with Fortune published the same day, said he would delay OpenAI's initial public offering, which had been planned for this year, citing safety concerns.
Nvidia CEO Jensen Huang, meanwhile, pointed to cybersecurity as the next market for AI. According to Bloomberg, Huang said at a Goldman Sachs Group technology conference in San Francisco that cybersecurity is likely to become the next major application area for AI. As advances in AI models accelerate the automation of computer programming, he said, both the exploitation of security vulnerabilities and the speed of patches that block them are picking up sharply, reshaping the cybersecurity industry as a whole.
