YouTube labels, removes, or de-amplifies content based on the nature and impact of misinformation, the online video-sharing platform’s director and global head of responsibility Timothy Katz has said. Any content violating policies, irrespective of how it was created, will be removed from the platform, he added.

“All content identified as misinformation may or may not violate our content policies. … If you want to, say, upload a video that no human has ever landed on the moon before, we think that is probably a video that should be allowed to be on YouTube because there is no risk of egregious real-world harm… But that does not mean we want to recommend that content. …But there might be … misinformation about when or how to vote that does have [an] egregious role and consequences. That would be something that we would want to remove from the platform. …it depends on the nature of the type of content.”

YouTube’s guidelines on misinformation prohibit users from posting “technically manipulated or doctored” content that “misleads users and may pose a serious risk of egregious harm”. Google’s ad policy prohibits manipulated media, deepfakes, and other altered content meant to deceive, defraud, or mislead users.

Deepfakes and AI-generated content

YouTube (India) director Ishan John Chatterjee that none of their key stakeholders want to be associated with a platform allowing fake news misinformation or deepfakes. “Our incentives are aligned with the governments and key stakeholders that we have to address this.”

Katz said while Google and YouTube have been using AI for a long time, especially to moderate content at scale, generative AI has changed a few things. “The barriers to entry for producing content drops significantly [with generative AI],” he said.

Katz said AI is a great way for them to train models for better content moderation. “The more content we see and when we see folks trying to manipulate different types of media, etc, it makes our systems better and stronger.”

It is now obligatory for all creators to disclose when they have created any altered or synthetic content that is realistic, including using AI tools, especially for sensitive topics such as elections, conflicts, and public health crises, or public officials. Creators who consistently default on this stand the risk of having their content removed, suspension from the YouTube Partner Programme (and thus loss of revenue), etc.

Advertisers posting election ads need to disclose if they include digitally altered or generated material.

YouTube will start labelling videos that have either been altered or are synthetic. Usually, this label will be in the description box. For more sensitive topics, there will be a “more prominent label” in the video player.

Content created using YouTube’s generative AI products and features will always be labelled as altered or synthetic. All images generated using Google or YouTube’s tools will have labels in the metadata identifying them as synthetically generated. They will be embedded with a watermark.

It is unclear if this watermark will be used for content generated using tools other than Google’s, or when the watermarked content was further manipulated to evade detection through watermarking.

Katz said this is new for Google. He added it is somewhat similar to Content ID, an automated identification system used by copyright owners on YouTube to identify their content even if their digital fingerprints or watermarks may have been altered. The BBC used this tool to get the controversial documentary on Prime Minister Narendra Modi removed from the platform this year.

Kartz said they do not expect AI-generated news content and the use of AI anchors to have any impact in terms of recommendations or monetisation as long as they adhere to their guidelines. “If it is high quality and coming from those [authoritative] sources, we would still allow that content to be monetised. We encourage creators and partners to lean into the space if they think it can help to serve our users.”

Chatterjee and Katz said that news consumption in India is very high on YouTube Shorts and through connected TV, especially live content. “We have seen a lot of growth, particularly during sensitive moments. With large news stories occurring such as during an election cycle, we just see very, very large growth from many of our authoritative partners across the board,” Katz said. “What has been encouraging is that we have seen different types of news partners growing on the platform, whether independent journalists, broadcasters, print publishers, digital native publishers, etc. We have seen growth across all of those different segments.”