A fabricated image of College Park, Georgia mayor Bianca Motley Broom kissing a man who was not her husband spread across Facebook, seen by thousands before corrections caught up. The image required no Photoshop skill — just a prompt typed into an AI image generator. Six years ago, the same attack would have demanded at least rudimentary image-editing ability. The barrier to producing convincing political disinformation has collapsed to near zero. The pattern is already repeating across levels of government. The White House's official X account posted a digitally altered image of activist Nekima Levy Armstrong that darkened her skin and added tears to her face; a deputy communications director later called it a "meme." A parody account circulated AI-generated images falsely placing New York mayor Zohran Mamdani and his mother with Jeffrey Epstein. Senator Raphael Warnock has documented deepfakes depicting him with fabricated health problems and grandchildren he does not have. In each case, the fake reached thousands before fact-checks followed. Public alarm is running well ahead of legislative action. A mid-June poll by Rad Campaign and Women Who Tech found 71% of respondents worry about AI-generated deepfakes affecting elections, while two in three expressed concern about AI being used to sexualize women. Concern about deepfakes outstrips concern about AI-driven job losses by roughly three to one. The public has identified the threat; the question is whether institutions can respond. Federal efforts are gridlocked. Outgoing senator John Kennedy offered a bill requiring AI firms to install a "kill switch" for rogue AI; Senator Rand Paul blocked it, calling the effort premature. House minority leader Hakeem Jeffries called for urgent AI guardrails. But Republicans attempted to pass an amendment preempting states from regulating AI as part of the One Big Beautiful Bill last year — it failed 99-1 in the Senate, a rare bipartisan rejection. The federal government has produced no binding deepfake legislation. State-level action is uneven and fragmented. Tennessee's Elvis Act uses right-of-publicity protections to prohibit AI voice cloning and digital likeness replication. Most states now require disclaimers when political campaigns use AI to manipulate candidate images. Georgia — where the Motley Broom deepfake originated — is not among them. Georgia legislators debated deepfake bills in 2024 and 2025 but enacted nothing. The structural problem is asymmetric speed. Generating a convincing deepfake takes seconds. Debunking it takes hours or days. Distribution on social media is instantaneous and algorithmic; corrections are slow and organic. Google's SynthID tool helped Reuters debunk the Mamdani images, but detection tools are reactive by design — they chase production, and production is accelerating. The window between a fake's release and its correction is the window in which elections can be manipulated. Motley Broom herself identified the constitutional tension at the center of the debate: balancing First Amendment protections against the right of individuals not to have fabricated accusations levied against them. "This is the tip of the iceberg," she said. The current legal regime offers defamation suits as the only recourse — slow, expensive, and available only after damage is done. The question is not whether AI deepfakes will be used to interfere in elections. They already are. The question is whether the legal and institutional response will arrive before the 2026 midterms or after.
