Facebook owner Meta on Friday announced major changes to its policies regarding digitally created and altered media ahead of an election that will test the company’s ability to police deceptive content generated by new artificial intelligence technology. did.
Starting in May, Mark Zuckerberg’s social media giant began applying a “Made with AI” label to AI-generated videos, images and audio posted to its platforms, and so far only a fraction of the videos have been manipulated. Monica, vice president of content policy, said that the policy, which previously only applied to departments, will be expanded. Bickert said in a blog post.
Bickert said that regardless of whether the content was created using AI or other tools, Meta is targeting digitally altered media that is “particularly at risk of materially deceiving the public about material matters.” also said it would apply a separate, more prominent label.
The new approach will change how the company handles manipulated content.
It will move from a focus on deleting a limited number of posts to one that maintains content while providing viewers with information about how it was created.
Meta previously announced a scheme that uses invisible markers embedded in files to detect images created by other companies’ generative AI tools, but did not reveal a launch date at the time.
A company spokesperson told Reuters the new labeling method will apply to content posted on Meta’s Facebook, Instagram and Threads services.
Different rules apply to other services such as WhatsApp and Quest virtual reality headsets.
Meta will begin applying a more prominent “high risk” label immediately, a spokesperson said.
The shift comes months before November’s presidential election, and technology researchers warn it could be transformed by new generative AI techniques.
Political campaigns in places like Indonesia have already begun deploying AI tools, pushing the boundaries of guidelines issued by providers such as Meta and generative AI market leader OpenAI.
In February, Meta’s oversight board reviewed a video of President Biden posted on Facebook last year and found that the actual footage had been altered to falsely suggest that President Biden had acted inappropriately. It criticized the company’s existing rules regarding manipulated media as “disjointed.”
Meta’s existing “manipulated media” policy only prohibits altered videos that are misleading if they are created by artificial intelligence or made to appear to say words that are not actually said. This footage was allowed to be released publicly.
The commission said the policy applies to non-AI content that is “not necessarily less likely to be misleading” than AI-generated content, as well as audio-only content or things that are not actually done. He said the law should also apply to videos depicting people in prison.
