Deepfakes, watermarking and content credentials
How AI-made images, voices and videos are spotted and labelled.
What is it?
Deepfakes are realistic images, videos or voice recordings made or changed with AI to show someone doing or saying something they never did. They can be used for fun and creativity, but also for scams, misinformation and harassment.
To help people tell what is real, companies and standards groups are adding signals to AI content. Watermarks are hidden marks inside an image, audio or text that software can detect. Content credentials are attached records that describe how a file was created and edited. Neither method is perfect, so careful checking still matters.
Which models use it?
- Image and video generators such as DALL-E, Midjourney, Stable Diffusion, Sora and Veo can produce very realistic media.
- Voice cloning tools can copy a person's voice from short recordings.
- Google DeepMind's SynthID adds invisible watermarks to some AI-generated images, audio, video and text.
- The C2PA standard (Content Credentials), backed by companies such as Adobe and Microsoft, records how media was made.
How to use it
- Pause before sharing shocking images, videos or audio, especially about public figures or breaking news.
- Look for the original source and check whether trusted news outlets report the same thing.
- Check for content credentials or AI labels on platforms and in tools that support them.
- Watch for signs such as odd lighting, mismatched lip movement or unnatural voice rhythm, but do not rely on them alone.
- Agree on a family code word to verify urgent calls asking for money, in case of voice-cloning scams.
- Label your own AI-generated media clearly when you share it.
Benefits
- Checking habits protect you from scams and misinformation.
- Labels and credentials help honest creators show how their work was made.
- Platforms can use watermarks and credentials to flag AI content at scale.
Limits and risks
- Watermarks and credentials can be removed or lost when files are edited, screenshotted or re-uploaded.
- Not all AI tools add watermarks or credentials.
- Automatic deepfake detectors make mistakes in both directions.
- Visual clues get harder to spot as generators improve.
Prompt template
Copy it, replace the parts in [brackets] and paste it into any AI chat.
Help me check whether this media might be AI-generated or manipulated. Description of the media: [what it shows]. Where I saw it: [platform/account]. Claimed source and date: [details]. Give me a step-by-step verification plan: what to search, which original sources to look for, what visual or audio clues to note, and how confident I can be. Do not give a verdict without evidence.
More in Safety and society
- Hallucinations and fact-checkingWhy AI sometimes states false things with confidence, and how to check its answers.
- Bias and fairnessHow AI can repeat unfair patterns from its data, and what you can do about it.
- Privacy and your dataWhat happens to what you type into AI tools, and how to protect personal information.
- Prompt injection and AI securityHow hidden instructions in content can trick AI tools, and how to stay safe.
- Copyright and AI contentWho owns AI-made work, and why the rules are still being decided.
- AI safety and responsible useSimple habits for using AI honestly, safely and with good judgement.