The Federal Reserve and the U.K. Prudential Regulation Authority have opened formal inquiries into major banks’ bank exposure AI hedge fund risk after the July 2026 collapse of Situational Awareness, an AI‑focused hedge fund linked to Jane Street and Citadel.
Situational Awareness lost 78 percent of its capital, wiping out roughly $15 billion for Jane Street and $10 billion for a private‑holdings vehicle. Within 24 hours, Citadel bought the fund’s public book at a 10‑percent discount. The rapid fallout prompted regulators to examine the financing web that connects banks, market‑making firms and AI developers, and to assess potential systemic risk.
In August, the Fed and the Bank of England sent questionnaires to banks requesting details on prime‑broker relationships with firms such as Jane Street, Citadel, Susquehanna and Hudson River Trading. The probes seek information on exposure size, risk‑appetite thresholds and intraday margin monitoring practices.
At the same time, internal Microsoft and OpenAI documents disclosed in a September lawsuit describe an AI‑driven “doom loop” that could erode website traffic and threaten publishers’ revenue. A separate multidistrict litigation in New York consolidates copyright claims from The New York Times, Mother Jones and other outlets alleging that OpenAI and Microsoft used copyrighted material without permission during model training. The plaintiffs argue the practice violates fair‑use standards.
Nvidia CEO Jensen Huang, speaking on The Ezra Klein Show, framed the issue as a product‑liability problem, saying any AI lab that cannot contain its models should be shut down. Huang’s comments suggest industry self‑regulation could be preferred to formal oversight, though no policy changes have followed.
Goldman Sachs, the fund’s largest prime‑broker client, earned more than $200 million in fees from lending to Situational Awareness during 2024‑2025, illustrating the scale of bank involvement. Regulators warn that if banks’ exposures exceed capital buffers, taxpayers could ultimately bear losses. Tighter capital constraints on market‑making firms could also reduce equity‑market liquidity and impair price discovery.
Publishers face a related risk. The alleged “doom loop” implies AI‑generated content may replace original reporting, cutting clicks and ad revenue. While the internal documents have not been independently verified, the lawsuit’s plaintiffs maintain the practice infringes on copyright. A recent $1.5 billion settlement with Anthropic shows courts may hold AI developers liable for using copyrighted material without permission, potentially reshaping AI‑training economics.
Huang’s product‑liability framing could influence future regulatory approaches. If unsafe AI labs are treated as liable entities, the cost of compute for labs that must demonstrate robust containment could rise, slowing development for smaller players.
Regulators have not disclosed the full scope of their investigations, and the primary source for the story—ZeroHedge—has a reputation for sensationalism. Nonetheless, the documented facts point to a hidden nexus that, if left unchecked, could expose taxpayers, media businesses and the broader economy to new risks. The next step will be the regulators’ findings, which could trigger tighter capital rules for banks, new liability standards for AI developers, and coordinated action from publishers seeking protection against AI‑generated content. The outcome will determine whether the market‑making ecosystem can adapt to AI‑driven pressures without compromising financial stability or intellectual‑property rights.