Yes. AI tools can produce realistic images, cloned voices and video of people saying things they never said. A political deepfake is one made to change how people vote. The best checks are not visual any more: look for the original source, other reporting, and an official account confirming it.
Yes. AI tools can produce realistic images, cloned voices and video of people saying things they never said. A political deepfake is one made to change how people vote. The best checks are not visual any more: look for the original source, other reporting, and an official account confirming it.
AI fakes are now cheap and fast to make.
Voice cloning needs only a short audio clip.
Looking closely is no longer a reliable test.
Trace it back to the source instead.
A deepfake uses AI to put someone's face or voice into content they were never part of. Early ones had obvious mistakes with hands and teeth. Newer ones often do not.
Verification has shifted from looking at the file to finding where it came from.
Search for the same clip on the person's official account or a major newsroom.
Check the date, and whether the event described actually happened.
Look for the full clip. Deepfakes are usually short and cropped.
Ask why it arrived on your feed right now.
Illustrations, diagrams and explainer graphics are often AI-assisted and perfectly honest, including some on this site. The problem is not the tool, it is pretending something happened when it did not.
Are political deepfakes illegal in New Zealand?
There is no single deepfake law, but other rules such as defamation and electoral advertising rules can apply.
Can detection tools spot them?
Sometimes, but they are unreliable. Sourcing beats detection.