Meta’s new AI detection tool doesn’t quite work as advertised, according to a new report.
Meta debuted Muse Image, its first image generation model, earlier this week. as part of debutthe tech giant also announced that all images generated by the model will include an invisible watermark system called Content Seal. The company claims that this signal will remain intact even if the AI-generated image is “cropped, compressed, resized or screenshotted” by users. To help catch the Content Seal signal, Meta Muse also announced that it is looking at an AI detection tool to check if Image has created an image.
But in a report published on Friday, Reuters Journalists found that the AI detection tool failed to identify more than half of the images it produced after being hacked. In a first test, Reuters found that the tool correctly identified all 40 Muse Image-generated images as AI, but after those images were cropped to half or a third of their original size, the tool could only identify 55% as AI.
As generative AI tools get better at producing extraordinarily deep forgeries, so does detection a more difficult problem to solve. According to a cybersecurity firm DeepStrikeThe volume of AI-generated deep forgeries online has experienced an annual growth of approximately 900% from 2023 to 2025. However, detection capabilities did not fully develop in parallel with this increase in popularity. According to previous research, commercial AI detection tools still suffer from errors, while the average person’s ability to identify AI-generated content is no better than flipping a coin.
While not a real immediate success, that’s the gap that Meta is trying to address with its new Content Seal and its detection tool.
Muse Image and its accompanying products were meant to be a big step forward for Meta, which has been outpacing its AI rivals. Last year, Meta CEO Mark Zuckerberg decided there was no time like the present to try to catch up and announced a major AI turnaround effort. The turnaround plan includes multibillion-dollar investments in research and development and poaching top talent from competitors across the industry. artificial super intelligence.
An additional few billion dedicated to artificial intelligence and a few more restructurings then Meta presented the first big fruits of this labor in April Muse Sparkownership model, which he says he plans to open source in the future and has met with mixed reception. Another highlight was the debut of Muse Image this week.
But the image generator’s debut and accompanying tools have been mired in controversy, not just according to the Reuters report. It was Instagram users excited knowing that an artificial intelligence model can use photos from any public profile without asking for permission from the owner of that profile. That feature now deleted.
The company is now eyeing its next big generative AI debut: a video generator called Muse Video. Hopefully, the company can address any loopholes in its detection tools and adequately address users’ privacy concerns before this model drops.
