Earlier this year, on the eve of the Chicago mayoral election, a video of moderate Democratic candidate Paul Vallas surfaced online. The post, tweeted by the Chicago Lakefront News, lashed out at Chicago’s lawlessness, suggesting that there was a time when “no one paid attention” to police shootings. seemed to be
The video looked authentic, but it was widely shared and the Twitter account vanished two days before the video was posted, before the Valas campaign denounced it as an AI-generated fake. It’s impossible to say if Vallas’ defeat to progressive Brandon Johnson, a former teacher and union organizer, had any impact, but it will disrupt public debate during the upcoming presidential election. A glimpse of possible high-stakes AI deception. . And that raises an important question: how do platforms like Facebook and Twitter mitigate them?
It’s a daunting challenge. There are no real laws regulating the use of AI in political campaigns, and it’s up to the platforms to decide what deepfakes users see on their feeds, and most are now autonomous. I’m having a hard time dealing with regulatory methods. “These are threats to our very democracy,” said Hany Farid, professor of electrical engineering and computer science at the University of California, Berkeley. forbes. “I don’t think the platform is taking this seriously.”
Currently, most of the largest social media platforms do not have specific policies regarding political or other AI-generated content.
On Meta’s platforms Facebook and Instagram, when content is flagged as potentially misinformation, third-party fact-checkers scrutinize that content, regardless of whether the content has been manipulated. You are required to defraud any forged, manipulated, or converted audio, video, or photographs. Meta spokesman Kevin McAllister said it is a classic photoshopping and AI-generated tool. forbes.
Similarly, Reddit will continue to rely on its policies for content manipulation. This policy applies to “disinformation campaigns, falsified documents, and deepfakes intended to mislead.” YouTube will also remove election-related content that violates its misinformation policy. This policy expressly prohibits images that are technically manipulated to mislead users and pose serious risks or potential harm.
On Twitter, owner Elon Musk announced an update to the company’s synthetic and manipulated media policy in April, stating that tweets “may” be labeled if they contain misleading information. said it will continue to remove tweets that are harmful to individuals and communities. The policy states that fabricated media will be scrutinized more rigorously “through the use of artificial intelligence algorithms.”
So far, the leading social media company with a more comprehensive policy aimed at moderating AI-generated content is TikTok.
In March 2023, TikTok announced a new “Synthetic Media Policy,” requiring creators who publish realistic-looking scenes generated or altered by AI to clearly disclose their use of such technology. rice field. For election content, we ban all AI-generated images impersonating public figures for political endorsement purposes.
“These are threats to our very democracy. I doubt the platforms are taking this seriously.”
TikTok began removing content that didn’t follow the new rules in April. Content creators have the flexibility to publish disclaimers and include them in subtitles, captions or hashtags as long as they are not misleading. The platform is not yet automatically banning accounts sharing unapproved content, it’s just issuing warnings.
However, while this kind of transparency is a first step toward alerting users to the origin of content, it may not be enough to prevent the spread of misinformation. “Of course there are bad actors out there trying to circumvent policies and standards,” said Lenny DiResta, research manager at Stanford University’s Internet Observatory.
That’s why maintaining team integrity is key to addressing this challenge, Diresta argues. Twitter in particular could struggle in the next election. Since Elon Musk took office, the company has fired the entire team working on misinformation and terminated contracts with third-party content moderators.
The spread of disinformation has long been a problem for social media platforms, and the January 6th Capitol raid demonstrated how movements organized largely online are having a deadly impact on real life. certified. But the 2024 presidential election will be the first election where campaigns and their supporters will have access to powerful AI tools that can generate lifelike fake content in seconds.
“Disinformation continues. We’ve thrown jet fuel into that problem in the form of generative AI and deepfakes,” Farid said.
But experts caution against overstating how harmful AI-generated content could be in the next election.
“Giving the impression that disinformation campaigns using deepfakes will inevitably succeed, when in fact they are not, can undermine trust in democratic systems,” said Cyberof the Georgetown University Security Center. AI researcher Josh Goldstein said. emerging technology.
Beyond social media, search engines also need to defend AI-generated content.
Google stepped in to filter out manipulated content from the highlighted results of knowledge panels and featured snippets. When it comes to Google Ads, manipulated media is prohibited and advertisers must go through an identity verification process and disclose within their ads who is paying.
In May, Google CEO Sundar Pichai unveiled a new “About This Image” tool that reveals that an image found in a search was generated by AI. The new feature will also show when a particular image or other similar images were first indexed and when they were seen elsewhere online, including in news, social media and fact-checking sites.
Pichai also announced that Google will soon begin automatically watermarking images and videos created by its generative models, making it easier for users to identify synthetic content.
Google is not alone in adopting this technique. Watermarking is one of the core demands of the Content Authenticity Initiative, a coalition of over 200 media, digital, content, and technology organizations promoting the adoption of industry standards for content authenticity.
Taking it a step further, Farid introduced the idea of requiring content creators to include a sort of “nutrition label” that reveals how the content was created. For example, the caption may state that the accompanying video was recorded with his iPhone 14 and edited with Adobe’s Firefly, his AI image generator.
An open question is whether tech companies can successfully self-regulate, or whether governments need to intervene. “I wish I didn’t have to [governments]said Farid. “But we still need help.”
follow me twitter. Send me your safety tips.
