- A growing legal debate argues that large language model (LLM) outputs are not 'speech' under the First Amendment, allowing more regulation.
- The attribution problem: determining whether a human created, selected, or adopted AI-generated text is difficult, and existing tools cannot reliably sort human from machine expression.
- If human attribution is the deciding factor for First Amendment protection, enforcement would rely on proxies like identity verification, which burdens anonymity and disproportionately affects marginalized groups.
- Personhood credentials (e.g., biometric scans) prove a person exists but not authorship, and their use could lead to increased surveillance and reduced privacy.
- Rejecting a categorical no-speech rule allows regulation of AI without sacrificing First Amendment protections for human expression, shifting focus to justifying burdens on specific rights.