Cayman’s Department of Communications has issued guidance about how the public can recognise artificial intelligence “deepfake.”
“In a fast-changing online world, AI deepfake technology makes it harder to tell what is real and what is fake,” the guidance states. “Deepfake videos, images, and audio clips use artificial intelligence to create false news that spreads incorrect information quickly.
Learning how these AI-generated deepfakes work and how to spot them is important for staying safe online.”
The guidance states that deepfakes are AI-generated content made with machine learning.
“A deepfake AI generator can take a real person’s face and put it on someone else’s body.
It can also make people appear to say things they never actually said,” the guidance states, adding that deepfake videos are often used to spread fake news, impersonation scams, and misleading content on social media. “These deepfake videos use algorithms that study real videos and then create new, fake ones that look real.”
The AI-generated content can be used to fool the viewers in numerous ways, including making fake political videos to mislead voters before elections, by criminals trying to trick people out of their money or to make it look like celebrities are endorsing certain products to increase sales.
In addition to AI-generated deepfakes, the guidance also warned about “cheapfakes,” which are made by editing images or videos using graphic arts tools like Adobe Photoshop.
Deepfake video
The guidance states that deepfake news and fake video content can be hard to detect, but offered these tips to do so:
- Look for strange movements. AI-generated faces may not blink naturally, and expressions might seem stiff or unnatural.
- Watch for lip-syncing issues. In deepfake videos, a person’s face and expressions may not perfectly match the words being spoken.
- Listen for audio problems. AI-generated voices can have odd pauses, robotic tones, or unusual speech patterns.
- Check the source. Fake news detection starts with checking where the content came from. Be cautious if it is from an unknown website or social media account.
- Use tools. Companies like Adobe, Microsoft, and OpenAI are making deepfake detection tools to help spot false information. Some apps can also help with deepfake detection.
- Look for bad lighting and shadows. AI-generated images may have shadows or lighting that do not match the rest of the scene. This can be a clue that an image is fake.
Deepfake audio
The guidance also warns that criminals are increasingly using AI-powered voice cloning to impersonate loved ones in distress, tricking victims into sending money. For instance, in the Cayman Islands, there have been reports of individuals receiving urgent calls from family members, purportedly in emergencies, pleading for financial assistance.
The guidance offers the following tips to recognise deepfake audio:
- Pay attention to tone and rhythm. AI-generated voices can sound flat or have an unusual rhythm.
- Listen for background noise. If someone is speaking but there are no natural background sounds, it could be AI-generated.
- Watch for mispronunciations. AI-generated content may struggle with difficult words or names, leading to strange pronunciations.
- Compare with real recordings. If you are unsure, compare the suspicious audio with a confirmed real recording.
