Safeguard Your Privacy: Weigh the Risks of Granting AI Access to Your Personal Information

Artificial Intelligence is becoming a more prominent part of our everyday experiences, influencing everything from mobile apps and software to search engines and drive-throughs. The introduction of web browsers with integrated AI assistants and chatbots signifies a notable transformation in how we access online information, especially compared to just a few years ago.

Nevertheless, the use of AI often demands extensive access to personal data, framed as a necessary requirement. This level of data access raises concerns that should not become standard practice.

Recently, it was striking to encounter seemingly benign free applications, like a “flashlight” or “calculator,” requesting permissions to access your contacts, photos, and live location. Even if these apps do not genuinely need such information for their stated functions, they often seek it when there are opportunities for monetization.

In this context, today’s AI tools do not differ significantly.

Take, for instance, Perplexity’s new AI-powered web browser, Comet. Comet enables users to search for information using its built-in AI search engine while automating tasks like summarizing emails and calendar events.

A recent TechCrunch review revealed that when Perplexity requests access to a user’s Google Calendar, it asks for multiple permissions related to the user’s Google Account. These permissions can include managing drafts, sending emails, downloading contacts, editing events across all calendars, and even accessing a complete corporate employee directory.

While Perplexity asserts that much of this data is stored locally on users’ devices, it effectively grants the company the right to utilize personal information, including for improving its AI models for other users.

Perplexity is not alone in its demand for extensive data access. There is a rising trend among AI applications claiming they can save time, such as recording phone calls or meeting discussions, which often require access to live private conversations, calendars, contacts, and more. Meta is also investigating what data access its AI applications may necessitate, including photos that users have stored on their devices but haven’t uploaded yet.

Meredith Whittaker, president of Signal, recently compared using AI agents and assistants to “putting your brain in a jar.” She observed that while AI can be helpful for routine tasks like making restaurant reservations or purchasing concert tickets, it frequently requires access to your web browser (which may reveal saved passwords, bookmarks, and browsing history), a credit card for transactions, and your calendar for scheduling—occasionally even access to your contacts for group reservations.

Utilizing AI assistants that rely on your data poses significant security and privacy risks. Granting such access entails relinquishing control over sensitive information—emails, messages, and calendar events—all in the name of convenience that is claimed to save time. Or as Whittaker put it, it reduces the need for thoughtful consideration.

Additionally, by allowing the AI agent to operate on your behalf, you place a high level of trust in technologies that may err or present inaccurate information. Engaging with AI necessitates trust in profit-driven companies that rely on your data to refine their AI models. When inaccuracies arise (which is common), staff at these AI firms may have access to your private queries to resolve the issue.

From a security and privacy perspective, the clear cost-benefit analysis of linking AI to your sensitive data suggests that surrendering access to such information is unwise. Any AI application requesting extensive permissions should raise alarms, much like a flashlight app seeking continuous access to your location.

Considering the extensive data you provide to AI companies, take a moment to reflect on whether the benefits you gain truly justify this exchange.