On Wednesday, Meta (formerly known as Facebook) announced the launch of its latest AI-powered image identification tool, the “Segment Anything Model.” The tool, which is still in demo mode, allows users to identify specific items within an image with just a few clicks. Meta claims that the tool can take a photo and individually identify the pixels comprising everything in the picture, allowing one or more items to be separated from the rest of the image.

According to Meta, “segmentation,” which involves identifying which image pixels belong to an object, is a core task in computer vision used in a broad range of applications, from analyzing scientific imagery to editing photos. The company says that creating an accurate segmentation model for specific tasks requires specialized work by technical experts with access to AI training infrastructure and large volumes of carefully annotated data.

To achieve greater generalization than previous approaches, Meta says it collected an unprecedented dataset of over 11 million images for the Segment Anything AI system. According to Ross Girshick, a research scientist at Meta, the company did not restrict the types of objects it annotated, resulting in a model with impressive capabilities to handle types of images that were not seen during training, such as ego-centric images, microscopy, or underwater photos.

Despite Meta’s assurances that the tool is still in its research phase with no plans to use it in production, concerns have been raised about privacy implications. A principal attorney at Oak View Law Group, Lyle Solomon, notes that data collection must be done transparently and with the individual’s full consent under privacy laws. Using AI for facial recognition without express consent raises questions about potential privacy law violations. Additionally, companies should avoid sharing facial data with third parties unless the individual has consented, and any sharing must adhere to privacy law provisions.

Meta’s pivot to focus on other products, including artificial intelligence, occurred after the company laid off over 10,000 workers following the cancellation of its Instagram NFT project. Meta’s new product group is focused on generative AI, with the Segment Anything Model being one of its latest projects. The development of AI has raised concerns among global leaders, leading to open investigations into the technology and what it means for user privacy and safety.

Kristen Ruby, the president of social media and AI consultant firm Ruby Media Group, notes that many users do not understand how the process of data collection works and the potential long-term consequences of their face being used to train a machine learning model without their consent. She suggests that companies should include a machine learning clause that informs users how their data is being used and whether they can opt out of future training models. Currently, many companies have an opt-in default setting, but this may change to opt-out in the future. Girshick notes that Meta has employed various privacy-preserving techniques, such as blurring faces and other personally identifying information, and that users can report offensive content to the company, which will remove it from the dataset.

In conclusion, while Meta’s new Segment Anything Model represents a significant technological advancement, it raises concerns about privacy and user consent. As AI continues to evolve, companies must prioritize transparency and privacy to ensure that users are fully informed about how their data is being used and can make informed decisions about opting in or out of training models.