REAL Political Advertisements Act

The bill

The long title says the bill amends the Federal Election Campaign Act to require "a statement within the contents of the advertisements if generative AI was used to generate any image or video footage." That title matches, word for word, the long title of S.1596 from the 118th Congress — the "REAL Political Advertisements Act," the source of the name above. What the statement would say, how large, where, on whose duty, carrying what penalty — no source exists for any of it, because the text is not published. What follows is what the published versions of the same idea say, clearly labeled.

What you should be aware of

• The older draft's trigger is any use. S.1596 (introduced May 2023 by Sen. Klobuchar) required the statement whenever an ad's image or video "was generated in whole or in part with the use of" generative AI. Images and video only — no audio. That draft contained no penalty section at all.
• The newer drafts' trigger is substantial generation. S.3875 (118th, reported by Senate Rules on May 15, 2024) and H.R. 9922 (119th, introduced July 23, 2026, "AI Transparency in Elections Act of 2026") narrow it to media "created or materially altered using generative AI," and exempt edits that don't change what a reasonable person takes from the media — cosmetic adjustments, color, cropping, resizing.
• Who the duty falls on, and what it costs, in the newer drafts: whoever pays for the ad. The payor carries the statement in a fixed format — letters at least as large as most of the text (image); at least four seconds, at the beginning or end, with a fixed script (audio); both (video); "permanently affixed." Enforcement: the FEC, through FECA's administrative-fine program — civil penalty up to $50,000 per covered communication, scaled by prior violations and reach; failure to answer an FEC complaint counts as an admission; the compliant statement itself is the safe harbor. Under the older draft: no penalty anywhere — an ordinary FEC fraud complaint is the whole regime.
• The enforcer's own record. In September 2024 the FEC declined to write AI-specific rules, called the existing fraud ban technology-neutral, and its majority said the agency "lacks legislative authority" for anything broader — while separately asking Congress to widen the ban. Today, the only operative federal rule on AI in campaign ads is that technology-neutral fraud provision.
• What courts have done with this design. Disclosure-only requirements inside campaign-finance machinery keep winning (Buckley; Citizens United; No on E v. Chiu). Laws that trigger on deceptive content rather than campaign-finance status keep losing — enjoined or partially enjoined in three 2025–26 cases: Kohls v. Bonta (California's AB 2839), Babylon Bee v. Lopez (voiding Hawaii's Act 191), and Accountability v. Knudsen (Montana's SB 25, over complaints against a watchdog's mailers). Which half of that record this bill joins depends on the one line nobody can read yet: the trigger.
• The permanence problem. The family drafts require the statement "permanently affixed" to the ad, and Pennsylvania's live law — effective December 21, 2026 — requires disclosure "permanent or extraordinarily difficult to remove … to the extent technically feasible." Platform pipelines strip C2PA content credentials from media at 0% survival on Instagram, X, TikTok, WhatsApp, and LinkedIn in one self-published test, and peer-reviewed watermarking research documents both removal and forgery attacks. The duty is priced onto a medium that currently defeats it. (Pennsylvania's line, for comparison: attorney general and district attorneys; civil fines up to $3,000 plus fees; no private right of action.)