On Monday, Meta, the social media giant, revealed a significant change in its AI labelling system, shifting from the “Made with AI” tag to the more nuanced “AI Info.”
This update comes after numerous photographers, including former White House photographer Pete Souza, raised concerns about real-life images being incorrectly labeled as AI-generated.
“As we’ve said from the beginning, we’re consistently improving our AI products, and we are working closely with our industry partners on our approach to AI labelling,” said Meta spokesperson Kate McLaughlin.
This move highlights the ongoing challenges and evolving strategies for AI transparency on social media platforms.
Dive into how Meta is navigating these complexities and what it means for content creators and users alike.
This seems to be rooted in the metadata tools, such as Adobe Photoshop,p that are applied to images and how platforms read the results of that data. The new “AI info” is less provocative and incorporates a much larger application of AI in creating and editing content. Meta thinks that way, some users at least, will be more savvy when they see the final edit without getting panicked over changes that are as routine as the one just made.
Meta Expands AI Startup Program Across Europe in Partnership with Hugging Face and Scaleway
“Our labels weren’t matching what people expected and didn’t give enough context,” Meta admitted in an updated blog post. They pointed out that even minor AI-assisted edits, like basic retouching, triggered the “Made with AI” label, potentially misleading users about how much AI was involved.
Label Update: Not a permanent Solution
However, the label update will not solve all the problems. Meta’s AI detection system does not change, it still relies on industry-level indicators and metadata from tools that correspond to C2PA and IPTC standards. This means that even if a photo has gone through some limited interaction with the AI editing tools, the ”AI info” tag will still be added.
This label change is but one of the strategies that Meta has implemented regarding processed AI media. In April, the company said that it would diversify its labelling policy after consultations with its Oversight Board and large-scale polls.
Join Meta’s AI Startup Program: Boosting European Innovation with Open-Source Models
Meta’s change in strategy demonstrates the main issues social media companies encounter when technology in artificial intelligence increases. This is because AI is becoming more capable and accessible, and distinguishing between the content generated with AI help, created with AI, and created by AI is increasingly challenging.
Thus, even though the new label is more permissive, the nature of the initiative raises several questions on what extent the usage of AI should require the disclosure of the same. Content creators and users alike will have an eye on how the practical use of this will be rolled out.
As AI systems redefine the digital environment, Meta’s label update is part of a continuous discussion on whether and to what extent technology, companies, content producers and consumers have to impose communication and creativity boundaries of AI on social media platforms. The new label may be helpful, but it is evident that the inhibition of the utilization of ethical and transparent AI on social media is still widespread.