Meta Offers Industry Standards to Detect AI-Generated Content

The Facebook and Instagram parent wants tech, media industries to align efforts to identify fake content ahead of the election

Smartphone display with logo of Facebook, WhatsApp and Instagram apps in hand against blurred META logotype on white monitor background: Tallinn, Estonia – October 29, 2021

Meta Platforms on Tuesday posted a list of standards for companies to use that can recognize photo, video and audio materials generated by artificial intelligence, offering the first large-scale effort to flag such content for audiences.

The post notes that polls show “a third of Americans say AI will reduce their trust in the results of the 2024 U.S. election,” and attempts to address the questions of how everyday citizens can know if the media they’re seeing is misleading and what organizations can do to improve transparency.

“There is no perfect answer,” the post states. “However, those building, creating, and distributing synthetic media can act to support better explanations of how content has been created and edited. They have a responsibility, when appropriate, to make clear to audiences that content is AI-generated or AI-modified (and any uncertainty about such judgments).”

“The need for alignment is urgent,” it continues, noting that policymakers in Washington are looking at what whether and how to regulate Ai and that media organizations are looking to “provide audiences with greater insight into what they see and hear.”

The standards Meta posted include a glossary defining the common technical methods that can provide insight into whether media is generated by AI or not; proposed questions to evaluate methods for disclosing the use of the technology such as watermarks and meta data, and guidance for policymakers and those building, creating, and distributing AI-generated content on how the techniques should be implemented.

If adopted by a range of companies, the standards could help social media platforms and other companies quickly identify posts that use AI-generated content and enable them to label it.

“While this is not a perfect answer, we did not want to let perfect be the enemy of the good,” Nick Clegg, Meta’s president of global affairs, told The New York Times. He said he hoped the effort would serve to rally tech and media companies to adopt a system for detecting and identifying AI-generated content so that it is easier for everyone to recognize it, The Times reported.

The move comes less than a week after Meta CEO Mark Zuckerberg apologized during a Senate hearing to the families of children harmed online.

The AI-generated deep fakes that have garnered the most attention in recent weeks are likely the ones featuring Taylor Swift and the late comedian George Carlin that surfaced last month. But The Times noted rising fears the tools will be widely used to produce fake content to sway voters, pointing to fake videos of President Biden circulating and a New Hampshire investigation of a series of robocalls that seemed to use an AI-generated voice that sounded like Biden to urge people not to vote in the recent primary.

Meta, the parent of Facebook, Instagram and What’sApp, is betting big on AI for its future, but has banned use of the technology to create political ads on its platforms ahead of the election, and requires other AI-generated content to be labeled as such.

The standards it’s proposing align with an effort by Adobe, maker of Photoshop editing software, and other tech companies are pushing the industry to adopt, The Times reported.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.