GSMA THRIVE NORTH AMERICA: Security company McAfee warned of a dark side to AI, with the technology increasingly employed to create deepfake content spanning social media, websites, video and even text.

Keynote speaker Celeste Fralick, chief data scientist and senior principal engineer (pictured), explained the public today faces a barrage of misinformation based on AI-manipulated content. Altered information is “very common throughout the internet”, with code widely available allowing anyone to “do a deepfake” by changing text, pictures and video.

This “low bar weaponisation of AI” is concerning, Fralick said, explaining malicious use of the underlying technology means “we must now all be sceptical of what we read, see and hear”.

She noted “deepfakes are a big business,” citing McKinsey Global Institute predictions of a market worth $2.6 trillion in sales and marketing alone using the approach, and noting the practice was increasingly being used to influence people on social media, particularly in political campaigns.

But social media giants including Twitter and Facebook are fighting back, separating the real from the fake “by using a combination of computer vision and deep neural networks”.

While there remain limits on what can currently be achieved, Fralick explained detection capabilities “are improving”, citing collaborative efforts including Reality Defender and Microsoft Visual Authentication.

Fralick also acknowledged there was a fun element to the use of AI to create deepfakes, citing face-swapping apps enabling people to put their own face to the stars of films or music videos.

Ultimately, she argued, “it is no longer the sole responsibility of technology or social media platforms” to flush out deepfakes, “it’s actually up to you to detect what is real and what is fake”.