Spotting AI-generated misinformation
Artificial intelligence tools can make synthetic photos, videos and audio nearly indistinguishable from real life, and such synthetic media may be deployed as part of disinformation campaigns to promote false information and manipulate audiences (CISA).
What official guidance advises
- Deepfakes are getting harder to detect. The Canadian Centre for Cyber Security says deepfakes, synthetic content including artificially generated images, audio and videos intended to deceive, are getting easier to make and harder to detect (Canadian Centre for Cyber Security).
- Slow down before you share. Don't immediately click to share content you see online; check the facts first. Some of the most damaging disinformation spreads rapidly via shared posts that seek to provoke an emotional reaction (CISA).
- Ask a few questions. Does it provoke an emotional response? Is it an extraordinary claim? Has it spread virally on unvetted or loosely vetted platforms? A yes does not automatically discredit the information, but it is a sign to research it further before trusting it (Canadian Centre for Cyber Security).
- Question the source. Research the author's credentials, consider the outlet's agenda and verify the supporting facts (CISA).
- See what credible sources say. Look at what credible sources are saying about the issue, and rely on credible sources of information such as government sites (CISA).
Practical checks
- Use a fact-checking site to see whether the information has already been proven false (Canadian Centre for Cyber Security).
- Do a reverse image search to check that images are not copied from a legitimate website or organization (Canadian Centre for Cyber Security).
- Don't assume that information is correct just because it comes from someone you trust, such as a friend or family member (Canadian Centre for Cyber Security).
- Check web addresses: the domain name of a fake site may have typos or use a different ending, such as .net or .org (Canadian Centre for Cyber Security).
- Look and listen closely: deepfakes often have lighting that doesn't match the background or other objects, and a voice that is slightly off or a mouth that does not match the words can be signs of deepfake audio manipulation (Western Australian Government).
- If you suspect a video is fake, report it to the platform you found it on (Western Australian Government).
- If friends or family share something that looks like disinformation, talk with them in a private, respectful conversation (CISA).
Sources
- Tactics of Disinformation, US Cybersecurity and Infrastructure Security Agency (CISA).
- How to identify misinformation, disinformation, and malinformation, Canadian Centre for Cyber Security.
- How to identify Deepfakes Cheat Sheet, Cyber Security Awareness Month 2024, published on wa.gov.au.