A deepfake label that vanishes on sharing is not enough.

Imagine receiving a political clip from someone you trust. You probably did not see its original upload, its first caption or the warning attached there. That ordinary act of sharing is why provenance matters beyond the platform screen where a piece of content began. The European Commission is consulting on revised election guidelines under the Digital Services Act. Its draft proposals address very large platforms and search engines: persistent labels for deepfakes, earlier detection of coordinated manipulation, testing of recommender systems and reporting after elections. The consultation opened November 10 and closes November 23; the page was updated November 12. These are proposed revisions, not newly adopted guidance or a fresh blanket ban. There is a constructive idea here: make the safeguards follow distribution, rather than judging safety only at upload. Yet a visible label is not proof that people notice it, understand it or change their judgment. Re-encoding, editing and movement between services also raise practical questions that a policy announcement alone cannot settle. The Commission says the proposals are content-neutral and do not request deletion of lawful content. That boundary matters. Platforms need transparent criteria and meaningful ways to challenge mistakes, especially when they intervene around candidates and political speech. A manipulation defense that cannot explain its own decisions could damage the trust it is supposed to protect. The evidence to watch is therefore operational: how much context survives sharing, how accurately coordinated abuse is distinguished from genuine participation, and whether published reports permit outside scrutiny. Better election protection should make information easier to assess—not ask voters to replace one opaque authority with another.