Meta has announced a new privacy feature for its upcoming AI-enabled smart glasses. Users will now be able to disable the use of any visual data captured by the device for training the company’s artificial-intelligence models. The change arrives amid growing criticism that Meta silently stores camera-based interactions and forwards them to overseas contractors for labeling.
When the opt-out is active, a request such as “translate this menu” will still be answered, but the snapshot used to generate the answer is immediately discarded and never enters the training pipeline. Meta emphasizes that the image will not be reviewed by humans, nor will it be saved to the user’s gallery.
What the opt-out actually covers
The setting applies solely to visual data – photos and video snippets taken while the glasses are engaged with Meta’s on-device assistant. Audio recordings remain subject to the company’s standard data-retention policies, although users can manually delete those files. Previously, Meta disclosed that images captured for AI tasks could be sent to third-party firms in other countries, where workers sometimes encounter personal or intimate scenes while labeling the content.
By offering a simple toggle, Meta hopes to mollify privacy advocates and address at least one pending class-action lawsuit that alleges the firm violated user consent. The move also coincides with the launch of a new Ray-Ban-branded line of glasses that omit cameras altogether, giving consumers a camera-free alternative.
Vatican adviser claims AI labs behave like a cartel
Father Paolo Benanti, the Vatican’s chief advisor on artificial intelligence, has publicly accused the world’s biggest AI laboratories – notably OpenAI, Anthropic and Google – of operating as a cartel. Speaking to Wired, Benanti argued that these firms repeatedly invoke the specter of super-intelligence to keep the real decisions about AI governance confined to their boardrooms.
According to Benanti, the focus on existential risk serves as a distraction, preventing regulators and the broader public from weighing in on the day-to-day choices that shape AI development. He is not denying that future dangers exist; rather, he contends that fear is being weaponized to maintain market dominance and to block competition.
Why the two stories intersect
Both Meta’s privacy toggle and the Vatican’s cartel accusation highlight a common tension: who gets to decide how powerful AI systems are trained and deployed. If the industry-led safety standards that companies like OpenAI and Anthropic are already drafting become de-facto regulations, smaller innovators could be priced out, and public oversight could stay minimal.
Recent legislative activity reflects this clash. Senators Josh Hawley and Ted Cruz stripped a national-security antitrust exemption for AI firms from the latest defense-authorization bill, while CEOs Sam Altman and Dario Amodei recently urged the United Nations to adopt global safety standards. The Vatican’s warning adds a moral dimension, suggesting that the current concentration of power may be less about safety and more about preserving control.
In short, Meta’s opt-out provision represents a concrete step toward giving individual users more agency over their data, whereas the Vatican’s critique calls for a systemic reshaping of AI governance to prevent a handful of corporations from writing the rules for the entire world.
