The Global Patchwork

The US regulates AI the way it regulates most things — sectorally. Executive Order 14110 (2023) briefly created reporting duties for frontier training runs before its revocation in 2025 shifted policy toward deregulation and procurement-led standards; the substantive machinery that persists lives elsewhere:

• Existing regulators — FTC polices deceptive AI claims and unfair data practices; FDA clears AI medical devices; EEOC has said employment anti-discrimination law applies to algorithmic tools (the first big AI hiring settlement came under existing law)
• States — Colorado's SB24-205 created consumer-protection duties for high-risk AI; California's AB 2013 mandates training-data transparency and SB 53 adds frontier-model safety/disclosure duties; dozens of state bills target deepfakes and bot disclosure
• Courts — litigation is de facto policy: copyright, defamation, and product-liability suits are writing rules faster than legislatures