Making Image Theft Harder: How Platforms Can Curb Deepfake Abuse

A face is no longer just a face. With a handful of public photos, freely available tools can now generate convincing video of a person saying things they never said. In Trinidad and Tobago we have already seen this used for fake investment endorsements, fake voice notes and intimate-image abuse.

Why it is so easy right now

Most social platforms serve images at full resolution, without watermarking, without provenance data, and without any friction around bulk downloading. A scraper can quietly collect hundreds of images of one person in minutes. The technology is not the problem, the defaults are.

What platforms should change

• Attach C2PA content-credential metadata to uploads so an image's origin can be verified later.
• Serve lower-resolution images to logged-out visitors and rate-limit bulk downloads.
• Apply invisible watermarks that survive cropping and re-compression.
• Give people a one-click, human-reviewed takedown path for synthetic media of themselves.
• Detect and label likely AI-generated media at upload, not weeks after it spreads.

What you can do this week

• Set older photo albums to friends-only; scrapers feed on public archives.
• Avoid posting high-resolution, straight-on, well-lit face shots, those are the most useful to a model.
• Reverse-image-search your own photos every few months.
• Agree a family safe word so a cloned voice on a phone call can be challenged.
• Screenshot, report and preserve evidence before an abusive post is deleted.

You cannot un-publish a face. But you can make the copy harder, slower and easier to disprove.

If a deepfake of you or your child is circulating, report it to the platform, preserve the evidence, and contact the TTPS Cybercrime Unit. Our Report a Cyber Crime page walks through exactly what to gather first.