For privacy and security, think twice before granting AI access to your personal data

Artificial Intelligence (AI) is increasingly integrated into various aspects of our daily lives—from smartphones and applications to search engines and even fast-food drive-throughs. The rise of web browsers featuring built-in AI assistants and chatbots signifies a fundamental shift in how we access and consume information compared to just a few years ago. However, many AI tools are requesting extensive access to personal data, often justifying this demand as a necessity for functionality. This level of access should not be taken lightly. Not too long ago, it was reasonable to question why a seemingly benign app, like a flashlight or calculator, would ask for permissions to access your contacts, photos, or real-time location. While these apps might not require such data to function, they often pursue access in hopes of monetizing it. AI applications are now following a similar trend. Consider Perplexity’s new AI-driven web browser, Comet. This browser offers users an AI-powered search engine and automates tasks, such as summarizing emails and calendar events. In a recent review, TechCrunch noted that when requesting access to a user’s Google Calendar, Comet sought a wide range of permissions. These included the ability to manage drafts, send emails, download contacts, and even access an entire employee directory from a company. Perplexity claims that much of this data remains stored locally on the user's device. Nonetheless, users are still granting the company rights to access and utilize their personal information, including for the improvement of its AI models for broader use. Perplexity is not the only company seeking such extensive permissions. A growing number of AI applications promise to save time by transcribing calls or meetings, yet they require access to very sensitive information, such as private conversations, calendars, and contacts. Even Meta is exploring the boundaries of permission requests, including accessing unuploaded photos on user devices. Meredith Whittaker, president of Signal, recently compared using AI assistants to “putting your brain in a jar.” She highlighted how these AI tools can take on tasks like restaurant reservations or concert ticket bookings, but to do so, they often request permission to access your browser, which could expose your stored passwords and browsing history, as well as your calendar and contacts. The implications for privacy and security are significant. By granting access, users are essentially surrendering a comprehensive snapshot of their most private information, including years of inbox messages and calendar entries, all in the name of convenience. This raises questions about whether the perceived efficiency is worth the risks. Moreover, allowing AI applications to act on your behalf demands a high level of trust in technology that is often flawed or can generate incorrect information. Users must also trust the profit-driven companies behind these AI products, which depend on user data to enhance their AI systems. When issues arise, it’s standard practice for AI firms to review user interactions to diagnose problems. From a security and privacy standpoint, conducting a simple cost-benefit analysis of linking AI to your private data can be alarming. Any AI application that requests extensive permissions should raise red flags, similar to an innocuous flashlight app wanting constant location access. Given the vast amount of data you may provide, it’s crucial to consider whether the benefits truly outweigh the risks.