Should AI-generated political ads be labeled?

In the 2024 cycle, the first major AI-generated campaign ads appeared. By 2026, synthetic media is cheap, fast, and increasingly realistic. The question of how to handle AI in political advertising is now urgent. Arguments for mandatory labeling: voters deserve to know when what they're seeing is fabricated. Even if the content is technically legal, deception is deception. Disclosures allow people to calibrate their trust. Arguments against: labeling requirements are technically difficult to enforce online, impossible to apply to foreign actors, and may simply shift to harder-to-detect methods. Some uses of AI in advertising are benign — not every synthetic ad is deceptive. The deeper question: is this a disclosure problem, a fraud problem, or a platform liability problem? Different frameworks lead to very different policy responses. If you were designing the rules from scratch — what principle would you use? And how would you enforce it against non-compliance, including from foreign actors who don't care about U.S. law?