In a significant move towards enhancing user experience while addressing privacy concerns, Microsoft has announced the launch of a new built-in feature called Recall. This innovative tool, which relies on artificial intelligence technology to capture snapshots of users' activities, is currently available in a preview version for users of Microsoft's portable computers under the Windows Insider program.
Recall, which Microsoft describes as a "personal co-pilot," aims to assist users in easily researching their previous activities, whether related to files, images, emails, or browsing history. For instance, if a user spotted a dress online a few days ago, they can utilize Recall to pinpoint its previous location.
However, the introduction of this feature isn't without its challenges. Microsoft initially halted the rollout of Recall in 2024 due to privacy concerns, with critics labeling it a "privacy nightmare." The company is now cautiously expanding the service's availability. European Union users, for instance, may have to wait until April 2025 to activate this feature.
Despite the improvements made regarding privacy, concerns remain. Privacy experts, who previously criticized Recall, have expressed apprehension about the potential for misuse. They warn that the feature could inadvertently capture information about other individuals without their consent, especially since it retains snapshots from messaging applications like WhatsApp, leading to the storage of images and messages that belong to others. One expert cautioned that if untrustworthy parties gain access to a device, they could exploit these captured images.
In response to these concerns, Microsoft has assured users that Recall does not share snapshots or data with the company or any third parties. This process requires user identity verification before activating the feature or accessing the snapshots. Additionally, Microsoft emphasizes that Recall only stores images locally on the device. Users can select which applications to capture snapshots from and can manually delete these snapshots when necessary, including avoiding capturing images while in private browsing mode.
On the regulatory front, the UK's Information Commissioner's Office (ICO) is closely monitoring Microsoft's progress with Recall. The ICO has underscored the importance of enhancing transparency and user data protection. While the agency does not grant pre-approval for products or services, it has reiterated that companies must comply with data protection laws and is prepared to take serious action if these companies fail to safeguard users' rights.
Meanwhile, Apple is also making strides in the realm of artificial intelligence, preparing to implement a new method for training its AI models that aims to enhance functionality while safeguarding user privacy. According to a report from Bloomberg and an official announcement on Apple's research website, this new technology is set to debut in beta versions of iOS 18.5 and macOS 15.5.
Currently, Apple relies on synthetic data—data generated artificially rather than collected from users—to train its AI features, including writing tools and message summarization. However, Apple acknowledges the limitations of this approach, particularly when it comes to understanding writing styles or summarizing lengthy messages.
To tackle this issue, Apple has developed a new method that compares artificial messages with real ones without accessing the user's email content. The company generates thousands of fake messages covering everyday topics, which are then transformed into what are known as "embeddings"—numerical representations of message content. These embeddings are sent to a select number of user devices that have opted into the device analysis program.
These devices compare the artificial messages with a small sample of real messages from the user and select the closest match based on content. Importantly, Apple ensures that these messages and comparisons never leave the device. Instead, anonymous signals are sent to Apple’s servers using a technique known as "differential privacy," allowing the company to analyze the artificially generated messages that are repeatedly chosen without knowing which device made the selection.
Apple has successfully implemented this privacy-focused approach in its Genmoji tool, which enables users to create custom emojis. The company tracks commonly used instructions anonymously to enhance the system's understanding of user preferences, while rare instructions remain invisible and are not linked to any specific user or device.
Furthermore, Apple has confirmed that it will apply these privacy-centric technologies to other AI tools soon, including Image Playground, Image Wand, memory creation features, and Visual Intelligence. This commitment to privacy reflects Apple's ongoing efforts to balance technological advancements with user rights and data protection.
As both Microsoft and Apple navigate the complex landscape of artificial intelligence and user privacy, they are setting new standards for how technology can enhance user experience without compromising personal data. With Microsoft’s Recall and Apple’s innovative training methods, users can expect more intuitive and privacy-conscious tools in the near future.
