Can AI be Consentful?

Consent alone cannot protect people from AI's unpredictable outputs: While people can agree to let companies use their data for AI training, they cannot meaningfully consent to all the different ways that data might be used to generate content about them—creating what the paper calls a "consent gap" that traditional privacy frameworks fail to address.

Three structural problems make meaningful consent impossible: The paper identifies the "scope problem" (we can't predict all possible AI outputs), the "temporality problem" (consent given today doesn't account for future uses), and the "autonomy trap" (people feel forced to consent to data use despite these unknowns), collectively making it impossible for individuals to truly understand what they're agreeing to.

AI-generated content creates new forms of personal representation beyond original consent: The paper highlights that AI systems can generate entirely new representations of people (such as deepfakes or synthetic media) that go far beyond the original data people consented to share, raising questions about identity rights and personal autonomy that current laws don't adequately protect.

Legal frameworks need fundamental redesign, not just tweaks: Current data protection and privacy laws are insufficient because they were built around traditional consent models; the paper argues that responsible AI requires evolving ethical and legal approaches that address fairness, transparency, accountability, and autonomy in ways that consent alone cannot achieve.