Facebook and Instagram will label all AI-generated images as fake

Meta plans to unveil technology capable of identifying and categorizing images produced by AI tools developed by other companies.

Meta plans to roll out its new technology across its platforms, including Facebook, Instagram, and Threads.

The company already marks AI-generated images from its own systems and aims for the new technology to encourage the industry to address AI manipulation, according to a blog post by senior executive Sir Nick Clegg. However, an AI expert cautioned that such tools are easily circumvented.

Clegg acknowledged in an interview with Reuters that the technology is still in development but emphasized Meta’s commitment to driving progress and encouraging industry-wide adoption in the months ahead.

Prof Soheil Feizi, who heads the Reliable AI Lab at the University of Maryland, pointed out that such a system could be circumvented easily.

According to him, “They might train their detector to identify certain images generated by specific models. However, these detectors can be easily bypassed with simple processing applied to the images, and they may also produce a high rate of false positives. Therefore, I don’t believe it’s feasible for a wide range of applications.”

Although Meta has acknowledged that its tool won’t be effective for audio and video, which are the primary concerns regarding AI-generated content, the company is urging users to label their own audio and video posts. They also stated they might enforce penalties if users fail to comply.

Sir Nick Clegg also acknowledged the impossibility of verifying text generated by tools like ChatGPT, stating, “That ship has sailed,” according to Reuters.

On Monday, Meta’s Oversight Board issued a critique of the company’s manipulated media policy, deeming it “incoherent, lacking persuasive justification, and inappropriately focused on content creation methods.”

Funded by Meta but independent of its operations, the Oversight Board’s criticism stemmed from a ruling regarding a video featuring US President Joe Biden. The video had been edited to depict Mr. Biden in a misleading manner alongside his granddaughter, suggesting inappropriate behavior.

Despite not violating Meta’s manipulated media policy due to its non-AI editing and portrayal of Mr. Biden’s actions rather than his words, the video remained on the platform.

While acknowledging the video’s compliance with Meta’s current fake media rules, the Board recommended updates to the policy.

Sir Nick, speaking to Reuters, expressed general agreement with the ruling and conceded that Meta’s current policy is inadequate for addressing the growing prevalence of synthetic and hybrid content.

Since January, Meta has enforced a policy requiring political advertisements to disclose the use of digitally altered imagery or video.

Disclaimer:
This content is AI-generated using IFTTT AI Content Creator. While we strive for accuracy, it’s a tool for rapid updates. We’re committed to filtering information, not reproducing or endorsing misinformation. – Jomotoday for more information visit privacy policy