Community Smart Hub · AI & You Series
Published 23 June 2026 · 5 min read · No tech background needed
Think of protecting yourself from fire. You need a smoke alarm, a fire extinguisher, a fire door, and a sprinkler system — not just one of them. This week’s research shows that protecting media from AI fakes works exactly the same way: you need multiple layers, not a single detector.
1. Why One Detector Is Never Enough
Experts who study deepfakes published a major review this week arguing that the era of relying on a single “Is this fake?” tool is over. As AI becomes more capable, a lone detector will eventually be fooled. The answer is to stack multiple lines of defence, each covering the weaknesses of the others.
Think of it like a bank vault: The cash is protected by a lock, a security guard, a camera, an alarm system, and a time-delay safe. Remove any one of these, and the others still hold. AI content verification needs to work the same way.
The Five-Layer Defence Against Fake MediaLayer
1: Cryptographic Provenance (C2PA Content Credentials)Digital birth certificate attached at creation. If the file is altered, the certificate breaks.Layer
2: Invisible Watermark (SynthID / SAiW)Hidden pattern inside pixels that survives cropping, compression, and screenshots.Layer
3: Content FingerprintA unique “digital fingerprint” of the content so copies can be traced even without metadata.Layer
4: Forensic DetectorAI-powered analysis looking for tell-tale signs of manipulation in pixels or audio.Layer
5: Human Review & GovernanceA person or team makes the final call. Rules and accountability sit behind all layers.✔ VERIFIED TRUSTWORTHY MEDIA
2. A Smarter Kind of Watermark That Tells You Who Made Something
Current watermarks just say “this was made by an AI.” New research proposes a smarter approach: a watermark that also tells you which specific AI system made it, and when — so if something goes wrong, you can trace it back to its source.
Source-Attributable Invisible Watermarking (SAiW) — A new type of hidden watermark that does two jobs: confirms an image was AI-generated, AND identifies which AI tool created it. Like a hallmark on gold that names the specific foundry, not just the metal.
The watermark is robust: it survives being compressed, cropped, filtered, or otherwise altered — the kinds of changes that commonly happen when content is shared on social media.
This shifts the key question from “Is this fake?” to “Who is responsible for this?” — which is far more useful for enforcement and accountability.
3. Two Technologies Working Together — Not Against Each Other
The industry has been debating whether to use C2PA certificates OR watermarks. The emerging answer is: both, because they protect against different threats.
- C2PA Content Credentials— Attached as metadata (hidden information linked to the file). Brilliant when preserved. But easily stripped when a file is downloaded and re-uploaded, or transcoded (converted from one format to another).
- Invisible Watermarks— Embedded inside the actual pixels or audio waves. Much harder to remove. Survives most platform transformations.
The EU’s own Code of Practice now recommends using both — with optional fingerprinting and generation logs on top.
4. The Gap Between Standards and Reality
Here is the honest truth: even when companies embed proper provenance information into content, many social media platforms strip it out when the file is uploaded. The label disappears before it ever reaches you.
Transcoding — The process of converting a file from one format to another. When you upload a video to social media, the platform usually re-compresses it. This often removes metadata — including content credentials.
The core challenge: Provenance tools only protect you if the platform you use bothers to keep and display them. Right now, most do not. This is the biggest barrier to progress.
5. What Good AI Governance Looks Like
A new academic review published this month concluded that no single detection tool will reliably catch all AI-generated content indefinitely. Instead, effective governance needs all of the following:
- Forensic detection— AI tools that look for manipulation.
- Cryptographic provenance— Digital certificates proving origin.
- Watermarking— Hidden markers surviving transformations.
- Regulatory frameworks— Laws that set minimum standards and enforce them.
- Organisational accountability— Companies being legally responsible for what their AI produces.
📌 What This Means for Our Community
The technology to protect us from AI fakes exists. The standards are being written. The laws are coming. The gap — right now — is implementation: getting platforms, governments, and organisations to actually deploy these systems and keep provenance information intact all the way from creator to reader.
At Community Smart Hub, we are working to make sure our community understands this gap — and can push for better.Community Smart Hub · AI & You Series · Blog 4 of 5 · Week of 20 June 2026
Sources: Scientific Reports (June 2026 review), SAiW research, EU Code of Practice, industry provenance analysis
