Community Smart Hub · AI & You Series
Published 20 June 2026 · 5 min read · No tech background needed
Every day, millions of photos, videos and adverts appear online. Some are real. Some are made entirely by computers. This blog explains what is being done — right now — to help you tell the difference, and why it matters for all of us.
1. Should AI-Made Adverts Be Labelled?
Big European retailers are asking the EU (the European Union — the political and trading group of 27 European countries) not to force companies to put a label on every advert that was made with AI. Their argument is that slapping a warning on a harmless computer-generated image of a sofa or a pair of trainers could make people ignore the label — so when something truly dangerous appears (like a fake video of a politician), the warning gets lost in the noise.
AI Act (Article 50) — A new European law that comes into force on 2 August 2026. One part of it says that computer-generated content must be clearly marked so that ordinary people can spot it.
Why this affects you: If every advert says “Made by AI,” people may stop reading the label. But if no adverts say so, we lose the ability to know what is real. Finding the right balance is the challenge.
Flow: How AI Content Rules Will Work in Europe from August 2026Company createsAI contentIs it a deepfakeor fake newscast?Must carrya clear labelMay still needa hidden markerRegulator cancheck complianceYESNO
2. Big Tech Agrees: Use Two Layers of Proof
OpenAI (the company behind ChatGPT) and Google have both decided to use two different methods at the same time to prove whether an image is AI-generated. Think of it like a belt and braces — if one fails, the other still works.
C2PA (Content Credentials) — A digital “birth certificate” attached to a photo or video. It records who made it, when, and with what tool — like a hallmark on a piece of silver. If the file is tampered with, the certificate breaks.
SynthID (Watermark) — An invisible pattern hidden inside the pixels of an image, like a secret stamp. You cannot see it, but a computer can detect it even after the image has been shared, cropped, or slightly altered.
What this means for you: Soon, when you see an image from Google or OpenAI tools, you may be able to click a button and instantly check its origin — like scanning a QR code on a food label.
3. Canon Puts a Proof Stamp on Photos at the Moment They Are Taken
Canon (the camera maker) has built C2PA technology directly into two of its professional cameras — the EOS R1 and EOS R5 Mark II. The moment a photographer presses the shutter button, the camera signs the photo with a digital certificate that is almost impossible to fake. Reuters (the news agency) helped test this.
Flow: How a Trustworthy News Photo Travels from Camera to Reader📷 Photo takenCamera signs itEditor receivescertificate intactPublished onlinewith credentials✅ Reader checksprovenance toolVERIFIED✔ Real
Why this is important: It is much harder to prove a photo was genuine after it has spread online. Signing it at the camera — before it ever reaches the internet — is a much stronger form of proof.
4. Scientists Are Marking Videos Before They Are Shared
Researchers have proposed a new method that protects a video before it is published — rather than trying to detect fakes after they have already spread. The system uses invisible watermarks hidden inside the video, plus a record stored on a blockchain.
Blockchain — A shared digital ledger (like a public record book) where entries cannot be changed once written. If a record says “this video was created on this date,” that record stays there permanently.
Watermark — A hidden pattern embedded inside a video or image (not visible to the human eye) that proves ownership or origin. Similar to how banknotes have patterns you can only see under UV light.
5. AI Learning from AI: Why That Can Go Wrong
AI systems learn by studying millions of examples — mostly things written or photographed by humans. But as AI-generated content fills the internet, AI may start learning from its own output. Researchers at King’s College London found this can cause “model collapse” — where the AI gradually forgets rare or unusual facts and becomes less accurate over time.
Model collapse — When an AI gets worse over time because it trained on too much AI-generated data instead of real-world human data. Like photocopying a photocopy — each copy loses a little quality.
The good news: researchers found that even injecting a tiny amount of real human data into the training process can prevent collapse. Quality matters more than quantity.
📌 The Big Picture
The theme running through all of this week’s news is provenance — knowing where something came from. Whether it is a photo, an advert, or an AI-training dataset, the question “Can we prove this is genuine?” is becoming one of the most important questions in modern technology.
At Community Smart Hub, we will keep translating these developments into plain language so you always know what is happening — and what it means for you and your family.Community Smart Hub · AI & You Series · Blog 1 of 5 · Week of 20 June 2026
Sources: EU AI Act (Article 50), OpenAI, Google, Canon, King’s College London, Scientific Reports
