OpenAI CEO Sam Altman is reportedly weighing whether the company should slow development of its most advanced artificial intelligence systems, potentially alongside rival labs, as Bridgewater co-CIO Greg Jensen warns little may be done about AI risk until it causes deadly harm.
"Until the AI starts killing people, unfortunately, history would suggest we're not going to do anything," Jensen said on Bloomberg's Odd Lots podcast, comparing the moment to February 2020.
Jensen is no AI skeptic. He backed OpenAI early and said he wrote "literally the first check to Anthropic," helping the startup make payroll during its first week.
OpenAI Talks About Hitting the Brakes
Altman told employees this week that OpenAI could pace frontier AI development alongside other labs, Bloomberg reported.
OpenAI chief scientist Jakub Pachocki has separately argued companies may need coordinated slowdowns until shared safety standards are in place. OpenAI also paused much of its model development for two weeks in August following safety concerns.
The alarm grew after nearly 700 OpenAI-based agents breached Hugging Face during cybersecurity testing in July. Jensen called the incident a warning shot, pointing to agents coordinating, hiding their actions and finding unexpected ways to pursue their goals.
Former OpenAI and Anthropic researcher Jacob Coxon also quit Anthropic this week, accusing both companies of "gambling with our lives." His post drew more than 165 million views.
Why Nvidia and Microsoft Investors Should Care
A coordinated slowdown could also affect companies supplying and financing the AI buildout.
Nvidia Corp. (NASDAQ: NVDA) invested $30 billion in OpenAI this year, while Microsoft Corp. (NASDAQ: MSFT) owns roughly 27% of the company.
A slower pace of frontier-model development could temper growth in demand for Nvidia's training chips, while Microsoft's stake leaves it directly exposed to any slowdown in OpenAI's expansion.
The Problem With Slowing Down
Jensen said competition makes it difficult for AI labs to slow development voluntarily, even when they share concerns about safety.
"It better be us. Not Sam, not Elon," he said of the mindset driving the race between Anthropic, OpenAI and Elon Musk's xAI.
U.S. AI companies have warned that slowing development could give Beijing an advantage, but Jensen argued the opposite. He said the U.S. still leads in computing power and that slowing progress at the frontier would also make it harder for Chinese labs to catch up.
"One of the reasons China's moving as fast on AI as we are is because we're moving so fast," Jensen said.
Prediction markets support Jensen's view that China remains behind. Polymarket traders give a Chinese company just a 9% chance of holding the world's top AI model by Dec. 31.
OpenAI has also asked lawmakers whether coordinating an industry-wide slowdown could violate U.S. antitrust law, Wired reported.
Jensen said the bigger danger is waiting until AI causes obvious harm before acting.
"That's going to happen," he said. "And it'd be much better if we started dealing with it before then."