Congress Must Take Action to Protect Against Growing AI Risks

Congress Must Take Action to Protect Against Growing AI Risks AI labs are building systems of immense power with little to no meaningful oversight.  There have been recent highly publicized AI security breaches that highlight the danger posed by Congress’ current stance: doing nothing, while the risk of serous AI-caused harm grows (notably, the same stance Congress is taking to the workforce and income issues virtually everyone agrees AI will cause).  This leaves every individual, business, and the U.S. government at risk.   One suggestion, proposed by Jen Easterly, a former Director of the Cybersecurity and Infrastructure Security Agency (CISA), is the formation of a federal AI risks council that brings together technical leaders from frontier AI labs (e.g., Anthropic, OpenAI) with the government’s top intelligence and security officials, to review incidents, near-misses, and newly discovered model capabilities — including intelligence on capabilities emerging from China and other foreign developers.   To my knowledge, no current federal legislation establishes this—or, any—kind of standing, recurring monitoring body. I’m asking you to:            1.     Direct the relevant committee(s) — Homeland Security, Commerce, and/or Intelligence — to hold a hearing specifically evaluating monitoring, and make proposals as to how protections around AI can be structured.          2.     Support legislation to create an interagency AI incident-monitoring mechanism.          3.     Make public your position on why Congress has not yet acted on AI oversight, given the specific warning bells we can all hear.   Congress does not need to wait for a catastrophic incident to justify a monitoring mechanism whose entire purpose is to prevent one.