Technology
Survey highlights privacy risks as users share personal secrets with AI chatbots
A study by DuckDuckGo Research indicates many users disclose sensitive information to chatbots while lacking awareness of data retention and legal risks.
The short version
- A DuckDuckGo Research survey found that roughly one-third of respondents share personal secrets with AI chatbots that they withhold from friends, family, and healthcare providers.
- Experts and researchers warn that chat logs can be incorporated into AI model training, accessed by employers, or subpoenaed by law enforcement.
- Only a quarter of survey respondents knew that mainstream AI companies are legally obligated to provide user chat data to the government or police under court order.
- Privacy advocates recommend using available opt-out settings, avoiding the disclosure of personal or company secrets, or utilizing local offline AI tools.
Key facts
- The DuckDuckGo survey found that 56% of self-described AI enthusiasts and 43% of parents share private information with chatbots that they do not tell trusted people in their lives.[CNET]
- Only 25% of survey respondents were aware that major AI providers like Anthropic, Google, and OpenAI can be compelled by law enforcement subpoenas to surrender user chat records.[CNET]
- Chat transcripts from AI services have previously been submitted as evidence in criminal legal proceedings, including murder and arson investigations.[CNET]
- A 2025 Stanford study identified six major tech firms—Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI—that used customer chat inputs to train their AI models.[CNET]
- Survey results showed that nearly 40% of participants distrusted private corporations to protect their data, while almost 50% lacked trust in the U.S. government.[CNET]
What remains uncertain
- The specific categories and exact sensitivity of the personal details shared by survey participants were not detailed in the DuckDuckGo report.[CNET]