Hany Farid, a digital forensics professor at the University of California at Berkeley who was not involved in Microsoft’s research, said if the industry adopted the company’s blueprint, it would become meaningfully harder to mislead the public with manipulated content. Sophisticated individuals and governments can try to circumvent such tools, he said, but the new standards could eliminate a significant portion of misleading content.
“I don’t think it will solve the problem, but I think it will solve a significant part of it,” he says.
Still, there are reasons to view Microsoft’s approach as an example of somewhat naive technological optimism. There is growing evidence that people are being swayed by AI-generated content, even when they know it’s false. And in a recent study of pro-Russian AI-generated videos about the Ukraine war, comments pointing out that the video was AI-generated received far fewer reactions than comments treating the video as real.
“No matter what you say, is anyone going to believe what you believe?” Farid asks. “Yes.” But, he added, “I think the vast majority of Americans and people around the world want to know the truth.”
That desire hasn’t led to urgent action by technology companies. Google began adding watermarks to content generated by its AI tools in 2023, which Farid said helped with the investigation. Some platforms use C2PA, a provenance standard that Microsoft helped launch in 2021. But Microsoft’s proposed set of changes, while powerful, could be just a proposal if they threaten the business models of AI companies and social media platforms.
