AI can help draft or summarise public comments, but MIT professor Deb Roy argues a person must remain answerable for the words. His proposed test for AI in civic life is not simply whether it can speak, but whether someone with a stake stands behind what it says.
Watch Desk analysis
What happened
In an interview with MIT News about his essay for the Bulletin of the Atomic Scientists, Roy warns that fluent AI-generated comments can be harder to distinguish from genuine public feedback. He points to the 2017 FCC net-neutrality consultation, where nearly 18 million of more than 22 million comments were later found to be fake, and argues that language models can remove the telltale repetition that once helped expose them.
Roy proposes four design principles: attributable authorship, processes that can be inspected, provenance built into the system, and widening who gets a say rather than concentrating influence. He argues that people can use AI to draft a comment, provided they adopt it and are ready to answer for it. Institutions could verify that a distinct, eligible person is speaking without necessarily making that person’s identity public.
Why it matters
Public consultation depends on officials and communities believing that comments reflect people with a genuine stake. If plausible machine-written submissions make that harder to judge, the damage is not limited to a crowded inbox: people may come to doubt whether anyone is listening, or whether a voice in the process belongs to a person at all.
Roy’s distinction is useful: AI assistance need not disqualify a contribution, but anonymous automation should not get to impersonate public participation. The awkward question is how to verify standing without exposing vulnerable contributors or handing a new gatekeeping power to institutions.
Our read
“Someone must answer for the words” is a stronger starting point than trying to ban AI from every civic conversation. It focuses on responsibility and makes room for useful tools, while insisting that participation cannot be reduced to a fluent text box. The hard part is building verification that earns trust rather than merely asking the public to trust the verifier.
What to watch
- Whether public consultation bodies introduce ways to verify eligible contributors while protecting their identities.
- How platforms disclose AI assistance and preserve a record of who adopted or edited a submission.
- Whether these safeguards widen participation or create new barriers to being heard.
Discussion spark: Should public-comment systems verify that each submission comes from an eligible person, even if that means adding friction or requiring a trusted intermediary?
Sources and evidence
- Q&A: Building accountability behind our words – and AI’s – in civic conversations (13 October 2026, 04:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.