Fact check: Monstrous sea creatures aren’t being pulled from the depths

Open this photo in gallery:

This article is part of The Globe’s initiative to cover disinformation and misinformation. E-mail us to share tips or feedback at disinfodesk@globeandmail.com.

Finned creatures with human features hauled aboard a trawler. Huge tentacled molluscs roaming the sea floor. A fisherman with a strange-looking white whale. None of the scenes in these videos are real, but they and similar ones have proliferated on platforms like Facebook, Instagram and TikTok where they can get tens of millions of views.

Tools such as Kling and Runway use generative artificial intelligence to make highly realistic videos with a few clicks, adding to the amount of “AI slop” – low-quality or misleading content created using AI – on the internet.

“It’s like spectacle on steroids,” said Clifton van der Linden, director of the Digital Society Lab and an associate professor in political science at McMaster University. “It’s creating something that is beyond the realm of reality but packaging it in such a way as to make one question whether it could be real.”

Spammers and scammers behind fake content

A study into these forms of fake content found that spammers and scammers were common culprits, seemingly motivated by profit or online clout.

Written by a researcher from the Stanford Internet Observatory and another from the Center for Security and Emerging Technology at Georgetown University, the study found some Facebook pages used fake content to attract users to then direct them to off-platform content farms and low-quality websites. Some of these sites were found to be selling dubious or non-existent products or attempted to obtain personal details.

“It could be that these were nefarious pages that were trying to build an audience and would later pivot to trying to sell goods or link to ad-laden websites or maybe even change their topics to something political altogether,” co-author Josh Goldstein told NPR. “But I suspect more likely, many of these pages were simply creators who realized it was a useful tactic for getting audience engagement.”

The Globe did not get a response after leaving a voicemail and messaging the phone number listed for a Facebook page that hosted about a dozen AI-generated videos of sea creatures. The Los Angeles address listed for the page does not match a real location.

Detecting fake videos a challenge

None of the videos on Facebook, Instagram, TikTok or X reviewed by The Globe had a label from the platform to indicate they were AI-generated or edited. Some creators self-disclosed the use of AI in their captions or used a hashtag to reference the generative tools they used.

The rapid improvement in the quality of AI video generators makes detecting fake videos a challenge for social media platforms, which handle vast amounts of content.

Ben Colman, co-founder and CEO of AI detection company Reality Defender, said via e-mail the issue of handling fake content is less about technology and more about the desire of platforms to combat the problem as well as the lack of any frameworks from governing bodies.

Meta (Facebook and Instagram’s parent company) said that image generators embed standardized indicators to show their AI origins, but that the same signals aren’t yet included in generated video or audio at the same scale. Images generated with Meta’s own AI have a visible watermark in the lower left corner.

Open this photo in gallery:

A TikTok blog post said the platform was implementing Content Credentials, a system to verify the authenticity and origin of digital content, to automatically label AI visuals. While TikTok responded to an initial e-mail, they did not give further information about content labelling by the time of writing.

An X blog post said that synthetic or manipulated media should not be posted that “may result in widespread confusion on public issues, impact public safety or cause serious harm.” X did not respond to an e-mail from The Globe seeking more information.

“We’re still in a period of trying to determine how to effectively communicate the authenticity of the content that is produced and disseminated on these platforms,” said Mr. van der Linden. “But there is also a deeper philosophical question of whether there is a moral or an ethical or some other kind of obligation to do so,” he said.

Tricks make videos seem more real

The videos we reviewed used techniques to make them seem more authentic, such as:

• Using a voiceover that sounds like renowned broadcaster and naturalist David Attenborough. A deepfake voice detector confirmed the sound-a-like voiceovers as fake
• Footage of genuine sea creatures edited together with clips generated by AI, blurring what is real and fake
• Some make bold claims such as saying that these creatures are new to science, have never been seen before, are caused by pollution or are evidence of a conspiracy, apparently to drive engagement

As detailed as AI-generated videos have become, there were several clues to indicate they were faked, including:

• Body parts blended together or appeared then disappeared
• The shape of objects like claws and teeth were inconsistent
• Abrupt changes in the speed objects moved
• Creatures appeared to be too light given their size when lifted
• Mispronunciation of places or names
• Missing or inconsistent natural sound. Indeed, many videos used the same piece of eerie music

The Globe also uploaded sample videos to an online AI detector, which returned high probabilities the clips were fake.

AI-generated video has come a long way. Can you spot the difference between real and fake? Take our quiz

Seas are home to genuine unusual creatures

There are ample strange-looking forms of life in the seas without resorting to AI to generate more. The Canadian Museum of Nature displays live sea cucumbers, anemones and urchins. Russian fisherman Roman Fedortsov posts images of real animals pulled from the depths to his social media, such as this fish with enormous eyes. Fair warning, some of his posts aren’t for the squeamish.