Proton has launched a new free tool called AI Paper Trail, designed to help users understand the privacy implications of their conversations with artificial intelligence assistants. The tool analyzes exported conversation data from platforms like ChatGPT or Claude to generate a personal privacy report, highlighting what personal information can be inferred from these interactions.
AI Paper Trail operates by analyzing the content of user prompts to identify various categories of personal data. According to Proton, the data uploaded for analysis is deleted immediately after the report is generated and is not stored on Lumo’s servers. The initiative aims to make the often abstract concept of AI privacy more tangible for users.
Over time, even seemingly innocuous individual questions posed to AI assistants can accumulate to reveal a significant amount of personal information. This can include details about a user's work, relationships, habits, interests, purchasing patterns, and travel plans. The tool demonstrates how a comprehensive profile can be constructed from a history of interactions.
When a user exports their conversation history, for example from ChatGPT, and uploads the resulting ZIP file to AI Paper Trail, the tool processes the most recent 200 prompts. It then produces a report that includes a privacy type, an AI Exposure Score, a summary of personal information revealed, and an estimated advertising value of that data.
One user's experience with the tool, analyzing their personal ChatGPT history, resulted in an AI Exposure Score of 58 out of 100, placing them in the "Leaving receipts" category. The report identified 47 distinct data points and estimated an advertising value of $185, along with five "red flags." The analysis highlighted exposure across categories such as location, interests, technology use, finances, and relationships, with location, interests, and technology showing the highest levels of inferred data.
The user noted that many prompts considered revealing were not individually perceived as sensitive. However, when combined with other conversations, they contributed to a much more detailed personal profile. While the inferences made by AI Paper Trail are not presented as definitive facts—for instance, researching a topic does not automatically mean it pertains directly to the user—the tool effectively illustrates how seemingly unrelated pieces of information can quickly build up to form a comprehensive digital footprint.
The launch of AI Paper Trail underscores the growing awareness of data privacy in the context of AI interactions. It serves as a practical demonstration of how extensive personal details can be reconstructed from the cumulative trail of prompts left behind in conversations with AI assistants, prompting users to reconsider the privacy implications of their digital dialogues.
