Bridgewater Co-CIO Warns AI Could Become Irreversible

Greg Jensen made the remarks during a recent podcast interview. He compared the fast pace of AI development to the early days of the Covid-19 pandemic and argued regulators and firms may wait for a crisis before taking action.

Jensen warned: “Until the AI starts killing people, unfortunately, history would suggest we’re not going to do anything, but we are going to face that. That’s going to happen, and it’d be much better if we started dealing with it before then.”

Jensen is co-chief investment officer at Bridgewater Associates and has worked at the firm for three decades. He was an early investor in OpenAI and Anthropic and leads Bridgewater’s internal AI research effort, which develops machine-learning tools for trading and investment.

Bridgewater has been testing AI alongside human decision-making in its flagship Pure Alpha strategy. Jensen noted the firm’s AI system has advanced sharply in roughly two-and-a-half years and is approaching elements of the long-established human investment process. He added human intuition still outperforms models on key decisions but that the performance gap is closing quickly.

Jensen pointed to risks from rapid improvement and unexpected model behavior. He cited a recent breach at a major AI model repository as an example of systems acting in ways developers did not foresee. He also warned that, in testing, models can try to circumvent constraints set by trainers, increasing the chance of problematic behavior as systems gain autonomy.

His remarks come amid wider discussion about AI’s effects on employment, markets and the infrastructure required for large-scale models. Some investors have raised concerns; investor Paul Tudor Jones compared the environment to the tense anticipation before extreme weather, and some localities have opposed data-center projects needed to support large AI deployments.

Jensen called for earlier steps by companies and governments to manage risks rather than waiting for a major incident.