Artificial Intelligence & Machine Learning
,
Next-Generation Technologies & Secure Development
Meta Changes AI Content Labelling, YouTube Updates Privacy Guidelines
Meta and YouTube updated their artificial intelligence policies to address the altered content appearing on their platform.
See Also: Close the Gapz in Your Security Strategy
Months after promising recourse for AI-generated content that impersonate users without their permission, YouTube is set to allow requests to remove the fabricated media, its updated privacy guidelines say.
The company says it will consider a “variety of factors” to determine if the reported content qualifies for removal, including depicting a “realistic altered or synthetic version of your likeness.” Another criteria for removal includes whether the content contains parody or satire. All complaints will be moderated by humans, with the video owner getting 48 hours to remove or cut out the offensive parts. YouTube does not specify how it will check any of these boxes at scale, as deepfakes continue to rise and large language models are notoriously lacking in determining context.
On YouTube’s criteria is also whether the content is disclosed to viewers as altered or synthetic.
Social media giant Meta also announced changes to how it labels content it suspects are AI generated.
Applicable to Facebook, Instagram, Threads and WhatsApp posts, the new label will display “AI Info” instead of “Made with AI.”
The changes come on the heels of accusations from artists that Meta’s detection systems labelled minor changes in images as being made with AI. This includes modifications such as cropping with an AI tool, which Meta detects as an AI modification.
The changes at YouTube and Meta come as part of industry effort to distinguish real content from fake, especially in the middle of an election year affecting countries across the globe.