The Clip Circulates by Sunday. The Rule to Judge It Doesn't Exist Yet.
Cloned candidate voices and AI campaign reels are already moving through barangay Facebook groups, and the job of spotting the fake lands on a first-time voter, not the platform.
A cloned voice endorsing a mayor, a reel of a senator saying something they never said, a clip stitched from real footage with a fabricated line dropped in. Content like this circulates in barangay Facebook groups less than two years out from the May 2028 vote, and it moves faster than any agency can write the rules to catch it.
COMELEC has issued rules on AI and deepfakes in past election cycles, and lawmakers have discussed proposals around synthetic political media. Whether the current framework covers the clips already spreading, and how it gets enforced this far ahead of a campaign period, is far less clear. The clips do not wait for a resolution. They land in a group chat named after a purok, get forwarded to a family thread, and reach a first-time voter before the enforcement question is settled.
Who gets stuck reading the fake
Here is the part that should bother you. When a deepfake spreads in your community group, the person expected to catch it is the 18-year-old scrolling at 11 p.m., not the company that runs the platform, and not the operator who paid to boost the post.
Platforms say they label some AI content and remove accounts that break their rules, but the detection load falls on the reader every time. You are asked to notice the mouth that does not sync, the audio that sounds a half-step off, the account created last month with a stock-photo profile. That is a lot of unpaid forensic work to demand from someone casting a ballot for the first time.
The tools are cheap and the reach is free
Voice cloning and video generation have gotten cheap and fast enough that convincing political clips no longer require a studio or a real budget, according to researchers who track synthetic media. The barrier to flooding a barangay feed keeps dropping.
Facebook remains where most of this lands, because that is where older relatives and neighbors actually are. The disinformation networks that shaped past Philippine elections did not disappear after the ballots were counted. They kept the pages, kept the follower bases, and now have generative tools that make the old copy-paste hoax look primitive.
The rule arrives after the vote is cast
Election bodies across Southeast Asia keep watching synthetic media outrun their timelines. Analysts who monitored recent national races in the region, including Indonesia's, have reported AI-generated candidate material in circulation, and the pattern tends to repeat: the clip spreads during the campaign, the response catches up after the results are certified.
Whatever COMELEC finalizes for 2028, the risk is that the clips arrive first and the enforcement arrives late. A takedown after the fact does not un-ring the bell for the aunt who already shared it three times.
So the ask stays the same, and it stays unfair. Verify the clip yourself, cross-check the account yourself, and correct the relative who believed it, all before you have finished your first year on the voter roll. The platform collects the ad revenue, the operator collects the reach, and you get handed the burden of being the fact-checker no one is paying.
