AI Misinformation Found Its Legislation. Now Two Coalitions Are Fighting Over What It's For.

Brazil announced this week that candidates who use AI to spread disinformation could lose their mandates. Manitoba introduced a parallel election misinformation bill. For a conversation that had spent months cycling through the same gallery of horrors — deepfake attack ads, synthetic Harris videos, a watchdog literally invoking Taylor Swift to signal how mainstream the threat had become — the arrival of actual legislation felt like oxygen. The dread didn't lift, but it organized itself around something actionable, and the energy across platforms shifted in a way that's hard to mistake for optimism but easy to mistake for progress.

It's worth pausing on what that shift actually represents. The people celebrating the legislative turn are, broadly, the electoral integrity community: researchers, journalists, policy-adjacent professionals who have been arguing for months that synthetic media is a democratic hazard and who now have Brazil and Manitoba as proof-of-concept. A study circulating among this crowd — examining how news audiences in Mexico, the US, and the UK differently process AI-generated misinformation — is getting real engagement precisely because it fits the new frame. The problem has structure; the structure suggests intervention; the intervention is underway. For this community, the narrative arc is satisfying.

For a different set of voices, it isn't. Advocates working on deepfake abuse have been pushing back, with increasing directness, against the tendency to treat electoral deepfakes as the paradigm case of the problem. The actual distribution of synthetic non-consensual content online skews overwhelmingly toward women and girls who are not politicians — private individuals with no institutional protection and no mandate to lose. The argument these advocates are making isn't that electoral integrity doesn't matter; it's that centering politicians in the deepfake policy conversation systematically buries the population most affected by it. When Manitoba passes a bill protecting candidates, it is not protecting the teenager whose face was put in a pornographic video. These coalitions share a diagnosis and diverge on whose harm counts as the organizing problem.

What the legislative turn revealed is that "pragmatic" isn't a single direction — it's a competition over which harms get codified first. Electoral integrity and survivor advocacy have coexisted in the AI misinformation conversation for months without forcing a confrontation, because dread is capacious enough to hold both. Rules aren't. The moment policy becomes concrete, prioritization becomes unavoidable, and right now the people with the most institutional access to the rulemaking process are the ones focused on elections. The survivors' advocates know this. That's why they're not celebrating Brazil.