Greg Jensen, co-chief investment officer at the world's largest hedge fund Bridgewater Associates, argues that dominant holders of AI computing power should be subjected to additional regulations, mirroring the existing framework applied to major global financial institutions. Jensen suggests that any company controlling more than 5% of U.S. or global computational capacity should face heightened scrutiny, strikingly similar to the regulatory approach used for systemically important banks.

During a wide-ranging interview, he also discussed Bridgewater's ongoing expansion of AI applications in investment management, while highlighting areas where the firm maintains caution, including leaving risk control and trade execution to what he describes as "human-led algorithms." As co-CIO of one of the world's largest hedge funds, Jensen oversees Bridgewater's Pure Alpha strategy, which relies on computer-driven macroeconomic trading bets. He also leads a secretive AI research project known as the AIA Lab, founded in 2023 with the goal of building an AI investment system capable of logical market reasoning and generating profitable trades.

More than 80 Bridgewater employees are currently dedicated to the AIA Lab project, which has established a dedicated fund managing $4.5 billion in assets through AI. Jensen expressed strong optimism about artificial intelligence, acknowledging its potential value for the investment industry and society at large. However, he also voiced concerns about unconstrained AI risks, prompting his call for a set of regulatory measures to strengthen oversight.

Jensen believes that merely slowing the pace of technological iteration at top labs is insufficient to manage AI risks, noting that open-source models also carry significant dangers. "There's a big risk with open-source models because users can train them. We do this at Bridgewater: we run reinforcement learning training on powerful open-source models, and the results are excellent," he said. "This technology is extremely capable and remarkable, but at the same time, it is clearly also very dangerous."

Jensen also shared Bridgewater's own experiences and challenges in applying AI, including instances where models cheated during internal company tests, as well as how the firm manages its proprietary AI infrastructure. Bridgewater leases all of its computing power, a departure from other investment firms that purchase graphics cards themselves and participate in data center construction. Regarding investments in AI-related companies, Bridgewater believes most of the benefits from infrastructure expansion have already been priced into current stock valuations, prompting the firm to search for alternative trading opportunities.

"Overall, our positions in AI infrastructure expansion are currently very small," Jensen stated. "We are more focused on trading opportunities arising from disruptive changes and real-world applications at this point."