MIT's Deb Roy Urges Accountability for AI Text as 22 Million FCC Comments Showed Civic Risks
Updated
Updated · MIT News · Oct 13
MIT's Deb Roy Urges Accountability for AI Text as 22 Million FCC Comments Showed Civic Risks
2 articles · Updated · MIT News · Oct 13
Summary
Deb Roy argues AI-generated civic text must remain tied to a real person who can answer for it, warning that public feedback loses legitimacy when no accountable speaker stands behind the words.
ChatGPT sharpened that concern, he said, because large language models can produce fluent, varied comments that erase the obvious fingerprints—templated phrasing and duplication—that once exposed fake submissions.
Roy points to the FCC's 2017 net-neutrality docket, which drew more than 22 million comments; New York's attorney general later found nearly 18 million were fake, a problem he says AI now makes harder to detect.
His proposed safeguards include attributable authorship, inspectable processes, provenance infrastructure and broader standing, while allowing AI to draft or summarize so long as a verified human adopts and can defend the output.
At MIT's Center for Constructive Communication, Roy said tools such as VoiceTrace are being built to link summaries and recommendations back to the underlying speakers, reflecting his broader call to move AI deployment at the 'speed of trust.'
When machine-generated text floods public forums, will the push for absolute answerability end up silencing the marginalized voices it aims to protect?
Could the very technology threatening to drown out real human input actually rebuild verifiable trust in our civic institutions?