COMELEC Wrote the Deepfake Rule for 2025. Nothing Confirms It Survives to 2028.
The poll body's 2024 AI resolution made deepfakes an election offense with jail time. It ran on administrative rules for one cycle, and no equivalent has been adopted for 2028.
COMELEC does have a track record on election deepfakes, and that is exactly the problem heading into 2028. On September 17, 2024, it adopted Resolution No. 11064, its guidelines on AI and social media for the May 2025 national and local races and the BARMM elections. The resolution required candidates to disclose AI use in campaign materials and let the poll body take down or block election content that used AI without that disclosure.
It also had teeth. Creating and spreading deepfakes, cheapfakes and softfakes, running fake or unregistered accounts, and coordinated inauthentic behavior all became election offenses under the Omnibus Election Code, carrying 1 to 6 years in prison with no probation.
A resolution is not a law, and it expires with the cycle
Those rules were an administrative resolution, built for one election, and poll watchdogs challenged them as possible censorship. No dedicated law on AI deepfakes in elections has been passed. It is not confirmed that equivalent rules have been adopted for 2028, which means the framework that governed 2025 is not guaranteed to govern the next mayoral bet whose face gets pasted onto a fabricated confession.
That is the gap. A fabricated clip can move through barangay group chats for days before any fact-check loads, and whether the poll body can order it down depends on whether it re-adopts and defends the same rules in time. An administrative resolution that has to survive a legal challenge is a thinner shield than a statute nobody can argue away.
Speed beats the correction, rule or no rule
Even with a resolution in force, the math favors the lie. A synthetic video needs one upload and a few reshares to reach a barangay-sized audience. A takedown needs a report, a review, an order, and then the same reach the original already had, which it almost never gets.
Platforms run their own labeling, but those systems apply unevenly and in English first, while the clips that decide a provincial race travel in Bisaya, Ilonggo, or Tagalog through closed group chats no reviewer can see. Facebook and TikTok both profit from the reach. A takedown order that pulls a clip from a public page does nothing for the hundred threads already forwarding it.
Foreign-run influence networks have already used regional platforms to seed content into Philippine feeds, and the tools to fake a face and a voice cost less now than the phones people watch them on. A local operator can commission a clip, and a coordinated network can amplify it, and both stay ahead of a review that runs on administrative timelines.
Watch what gets re-adopted, not what gets promised
The honest position is narrow. The 2025 rules showed COMELEC can name deepfakes as offenses and attach prison time. The open question is whether that carries into 2028 as a resolution that holds up in court, or whether Congress finally writes it into law so it does not reset every cycle.
A candidate can be smeared on a Tuesday and cleared on paper the following month, and still lose the seat the fabrication cost him. The penalty that would deter the operator only works if the rule exists before the clip goes up, and right now, for 2028, that rule is not confirmed to exist at all. The clock is already running.
