Image generated using Muse.AI

MANILA - YouTube is boosting its drive against deepfakes and implementing stricter policies on AI-generated content on the platform. 

During a virtual roundtable with journalists, YouTube officials said the platform is also tightening its rules on political ads as election campaigns move into high gear in several countries. 

YouTube officials acknowledged that AI-generated content, especially deepfakes, needed to be strictly monitored on the platform, as these may be used to influence elections in many countries. 

Alejandro Borgia, YouTube Director for Product Management, said many political parties and their supporters post ads on YouTube. 

“And in a major election year like this one for so many countries, we want to make sure that voters trust the election ads they see on our platforms. That's why we made verification and transparency foundational components of election advertising on Google,” Borgia said. 

OpenAI seeks to allay election meddling fears in blog post

He said that before an advertiser can run an election ad, YouTube first verifies that they are eligible to run election ads in their region or country. After this, YouTube verifies the advertiser's identity. 

“These two steps are critical for ensuring that bad actors don't use our ad products to influence or undermine our elections. Only after both these verification steps are complete is an advertiser able to run election ads,” Borgia said.

DEEPFAKES IN ELECTION CAMPAIGNS

Political ads that use AI-generated content to influence elections meanwhile present a new challenge, according to Duncan Lennox, YouTube Vice President and General Manager for Ads Privacy & Safety.

“We've been discussing the growing role that AI is playing in our industry and our society at large. And we fully recognize the impact that AI can have in politics,” Lennox said.

Of particular concern is the use of deepfakes and AI-generated videos and photos on political ads on YouTube. 

He said YouTube is making major investments to detect and block policy-violating AI-generated content as well as new transparency requirements for election ads. 

Borgia said the company launched a new policy last year requiring all verified election advertisers to disclose when their ads contain synthetic content, including AI-generated content. 

“So that people know what's real and what's not,” Borgia said. 

YouTube said it is launching a content-labeling tool within Creator Studio that requires creators to disclose when realistic content, which viewers could easily mistake for real, is produced using altered or synthetic media, including generative AI.

Some examples of content that require disclosure include:

• Using the realistic likeness of a person: Digitally altering content to replace the face of one individual with another's or synthetically generating a person’s voice to narrate a video.
• Altering footage of real events or places: Making things look different from how they are or happened, such as making it appear as if a real building caught fire, or altering a real cityscape.
• Generating realistic scenes: Showing a realistic depiction of fictional major events, like a tornado moving toward a real town.
  
  

“Labels indicating the use of altered or synthetic media will appear in both the expanded description and, for sensitive topics like health or news, prominently within the video player itself. These labels will gradually roll out across all YouTube platforms in the coming weeks, ensuring consistency and clarity for viewers,” the company said in a statement.

Lennox meanwhile said YouTube’s policy against using manipulated images in political campaigns predates the widespread use of generative AI.

“If you're attempting to deceive, to fraud or scam people, as in the use of a deepfake to, for example, in an election ad, have a politician appear to say something they never said, for example, then that would violate our manipulated media policy,” Lennox said.  

But he also acknowledged that generative AI is constantly evolving and that YouTube also needs to adjust its enforcement and detection capabilities to keep pace.

Deepfakes have been identified as an issue in India where political parties are harnessing powerful AI tools to make deepfakes, reproducing famous faces and voices in ways that often appear authentic.

Ghost in the machine: Deepfake tools warp India election

Both the government and campaigners have warned that the spread of such tools is a dangerous and growing threat to the integrity of elections in India.

Here in the Philippines, the Department of Information and Communications Technology has warned of a possible increase in deepfakes and other artificial intelligence-generated fake videos as the 2025 mid-term elections near.

DICT warns of possible rise in deepfakes as 2025 elections near 

Meta to require political ads reveal AI altered images

Lennox said that in 2023, YouTube verified more than 5,000 new election advertisers and removed some 7.3 million ads that came from advertisers who didn't complete verification.