The Fed's AI warning is not a prophecy of collapse.

AI does not need to become a single hostile intelligence to create a financial-stability problem. Many institutions can make individually reasonable decisions that leave them exposed to the same supplier, financing assumptions or operational shock. That is a more useful reading of the Federal Reserve's November Financial Stability Report than treating it as a prediction of an AI-triggered crash. In its market-contact survey, 87% of 23 respondents cited AI, compared with 50% of 20 contacts in the spring. These are small, different survey groups identifying risks over the next 12 to 18 months. They are not the percentage of banks controlled by AI, a representative public poll or an 87% probability of collapse. The report explicitly distinguishes respondents' views from those of the Federal Reserve Board and New York Fed. The concerns span valuations, financing complexity and cyber or operational damage. Those mechanisms can reinforce one another, but their appearance in a risk survey does not show that the chain reaction has begun. Nor does productive use of AI eliminate exposure to a shared service failure. The practical question is concentration that an institution can actually measure: which obligations depend on the same counterparty, which processes fail together, and what can continue if an AI service disappears? The answer may support continued investment with better limits. It may also reveal that apparent diversification is several interfaces resting on one dependency. Neither conclusion requires a movie-style takeover.