Meta, the parent company of Facebook, has announced plans to introduce new technology capable of detecting and labeling images generated by artificial intelligence (AI) tools developed by other companies. This technology will be implemented on Meta’s platforms, including Facebook, Instagram, and Threads.
Currently, Meta already labels AI-generated images produced by its own systems. The company aims to extend this labeling to images generated by external AI tools in the coming months, as stated in a blog post by senior executive Sir Nick Clegg.
In an interview with Reuters, Sir Nick Clegg acknowledged that the technology is still under development and not yet fully mature. However, Meta hopes to create momentum for the industry to address the issue of AI-generated fakery.
Despite Meta’s efforts, some experts remain skeptical about the effectiveness of such detection tools. Prof Soheil Feizi from the University of Maryland’s Reliable AI Lab expressed concerns about the ability of these detectors to be easily evaded and the potential for a high rate of false positives.
Meta has also admitted that its new tool will not be applicable to audio and video content, which are often the focus of concerns regarding AI-generated fakes. Instead, the company is relying on users to label their own audio and video posts and may impose penalties for non-compliance.
Regarding text generated by AI tools, such as ChatGPT, Sir Nick Clegg conceded that it would be impossible to test for manipulation.
This announcement comes in the wake of criticism from Meta’s Oversight Board regarding the company’s policy on manipulated media. The Board deemed Meta’s policy “incoherent” and called for updates to address the growing prevalence of synthetic and hybrid content.
Sir Nick Clegg agreed with the Oversight Board’s assessment, acknowledging that Meta’s current policy is inadequate in the face of evolving challenges posed by synthetic content.
Since January, Meta has implemented a policy requiring political advertisements to indicate the use of digitally altered images or video.