Data Brokers Sell Sensitive Chat Transcripts

🔑 Enhanced Key Takeaways

• •Browser extensions capture real-time AI conversations, store them verbatim in vector databases, and expose them via API using pseudonymized SHA-256 hash IDs.[1]
• •Healthcare workers are inputting actual patient data into chatbots, resulting in commercial databases containing medical record numbers and diagnosis codes.[1]
• •Koi Security reported in December 2025 that 8 million users' AI conversations were sold for profit by privacy-focused browser extensions.[1]

🛠️ Technical Deep Dive

• •Data captured via browser extensions from private AI conversations.[1]
• •Stored in vector databases for efficient searchability.[1]
• •Pseudonymized using SHA-256 hashes for user IDs, but conversation content remains verbatim and includes identifiable details like names, DOBs, and medical codes.[1]
• •Exposed to customers through authenticated API endpoints.[1]

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline