Federal plan to rein in biased police AI has loopholes
• The federal government's proposal to constrain AI use by law enforcement is commendable, but allowing waivers for "critical operations" could undermine protections.
• Facial recognition and other AI have already led to false arrests and harm, especially for Black individuals, due to biases and lack of safeguards.
• The proposal requires assessing benefits and risks, testing real-world impact, and getting community input before deploying AI tools.
• It bans non-compliant AI after August 2023, but allows individual agencies to seek waivers, which risks undoing protections.
• Federal guidelines should also apply to state and local law enforcement through tying compliance to federal funding.