MIT’s Deb Roy: We should be accountable for every text in the age of artificial intelligence

Summary

MIT professor Deb Roy argues in a Bulletin of the Atomic Scientists article that every piece of text in public life—whether human- or AI-generated—must be tied to a responsible human. Roy warns that large language models (LLMs) are eroding the legitimacy of democratic feedback channels, such as public comment periods.

Why it matters

Public comment processes are a cornerstone of democratic legitimacy. The article documents how AI-generated texts silently hollow out these processes by eliminating the detectable 'fingerprints' that expose mass fake feedback. Roy’s call for structural accountability is particularly timely, especially given reports that governments struggle to distinguish between genuine and synthetic comments.

Highlights

• In the 2017 FCC net neutrality case, 18 million of 22 million comments were identified as fake due to templated sentence patterns. LLMs eliminate this clue, making fraud detection significantly harder.
• Roy’s core philosophical claim is that even if we accept LLMs have full comprehension capabilities, they lack 'legal status.' The chain of responsibility cannot end with a software system that has nothing to lose.
• Roy’s own institution, the MIT Center for Civic Media, uses AI to theme comments but offers human intervention at every step. This practice serves as a concrete example of the accountability model Roy advocates.

Critical take

The article reads like a single-voice promotional Q&A section for Roy’s essay. The principle of accountability is presented as an unquestionable fact; no regulatory mechanism, platform policy, or legal framework is outlined for how accountability would be verified at scale. Additionally, the text cuts off mid-sentence, the Q&A section remains incomplete, and the strongest counterarguments (e.g., who is responsible when a user prompts an LLM or when the LLM’s output deviates from intent) are not addressed.