One of the biggest risks with AI is recursive self-improvement—where AI agents continuously code better versions of themselves that could evade human oversight.
I spoke to ABC News Live about one of the technically feasible ways Congress can address this.