You can’t believe anything on the internet

You can’t trust videos

A new technology has been developed using AI to find faces in videos and map a different face on top.
The software works by making a face map (Like the face map on snapchat when you activate the filter) of the original face and the new face, then manipulating the new face to stretch and rotate to match the facial movements and most importantly match the movement of the lips.
This is then repeated with slightly different permutations, some closer to realism and others even less human than before.
Working parallel to the first software is a second software that selects the ‘fittest’ results and terminates the others. (Survival of the fittest, but with videos.)
While this technology can be used to declare war on other nations as a figure of power, or used to frame someone for something they didn’t do, luckily people at the moment are preoccupied placing Nick Cage into random movies.

...And you can’t trust audio

Thanks to services like Lyrebird, given enough samples of a person’s voice, their technology can match the speech patterns so that any sentence can be read in their voice even if the specific word has not been heard by the system before.
Voice-cloning also has a similar issue to deepfakes where identity can be stolen. LyreBird is well aware of this and have raised awareness by faking audio of famous people warning of the dangers.
A really unexpected use for this is that with a vast voice pattern database, it is possible to find illnesses that affect the voice, but are too subtle to be heard by humans. For example Parkinson's disease can be detected by this system and as a result, we now know that Hitler had the disease. Obviously being able to find a disease from a voice recording has loads of potential that is yet to be used, as it is not used effectively in medical scenarios today :)