Representative Ro Khanna, a Silicon Valley Democrat, plans to introduce an AI safety bill that would prohibit "recursive" self-improving AI until the government establishes safeguards. The legislation also includes strict liability standards, making companies responsible for harm caused by their AI systems.
The bill arrives as top tech executives have warned that AI could become uncontrollable, and as Congress has largely failed to pass significant AI regulation despite the technology's growing role in the economy. The proposed ban targets AI that can improve itself in ways that may outpace human oversight.
Under the strict liability provisions, developers would face legal exposure for damages from their systems. The bill is part of a broader push to address risks that some industry leaders say could lead to catastrophic outcomes if left unregulated.