Meta changes ‘Made with AI’ label to ‘AI Insights’ to give more details about the content of images

Meta has begun to label images that have been generated or edited with Artificial intelligence (AI) as ‘AI Information’, rather than using the label ‘Made with AI’, to provide more details about how the technology has been used and whether it involves “minor modifications”.

To promote transparency in its tools, the technology company led by Mark Zuckerberg announced in February of this year that it would begin to identify realistic content generated by AI tools that was published on its social networks (Facebook, Instagram and Threads).

Currently, these labels are displayed as ‘Made with AI’ for images that have used this technology. However, they apply both to content that has actually been generated from scratch using AI, and to images that are real but have been retouched using some tool.

Some professional photographers have shared their complaints about this measure, as they claim to have shared photos with some retouching done by AI to clean up the image and, despite showing authentic content, they were automatically labeled ‘Made with AI’. This was explained by those affected at the end of June, among whom was the former White House photographer, Pete Souza.

Meta claimed that they were working on a new approach that would indicate the degree of AI tool use that a labeled content has. “We are taking into account recent feedback and continuing to evaluate our approach so that our labels reflect the amount of AI used in an image,” it explained then.

It has now begun replacing the ‘Made with AI’ label with ‘AI Insights’ for content that uses AI-powered tools, so users can tap on the label to learn more about how AI was used in the image.

The technology company has detailed this in an update to the labeling statement for AI-generated content on its website, where it has clarified that its labels “were not always aligned with people’s expectations and did not always provide sufficient context.”

The firm has said that to assess whether something has been created with AI, it relies on industry-standard indicators that “other companies include in content created with their tools.” If content posted on its social media channels falls within these standards, it is labeled “Made with AI.”

However, it has admitted that some of the content it flagged included only “minor modifications” made using AI tools, such as retouching the colours or focus of an image.

To clarify the reasons why content is labeled, Meta has updated the label to ‘AI Information’ across all of its social networks, so that users will only need to click on the label to receive information about why it has been labeled and to what extent AI has been used.