Hundreds of Contractors Reading Real ChatGPT Conversations
OpenAI is reportedly paying hundreds of contractors to review and rate ChatGPT responses under the codename Project Lily. This practice has raised concerns about user privacy, especially regarding data collection and informed consent.
In summary, OpenAI has not disclosed where it informs users that human reviewers might access their chat history. This issue was further complicated by the Court of Justice’s ruling in September 2026, which stated that the duty to inform users applies at the time of data collection, regardless of whether the recipient can identify individuals.
According to 404 Media, contractors do not see usernames during their reviews but may access user memory summaries, providing limited context about previous conversations and locations. OpenAI’s Privacy Filter model attempts to redact personal information, but it is not foolproof.
The European ruling in EDPS v SRB emphasizes that the data controller’s obligation to inform users takes precedence, regardless of the recipient’s ability to identify the source. This decision has significant implications for companies like OpenAI, especially with Italy’s fine of €15 million for non-compliance.
OpenAI’s privacy settings offer some control, but they vary across different user types. Free and Plus accounts have improved privacy models, while Enterprise, Business, and Edu users can opt for zero data retention.
A reviewer expressed skepticism about the implications: "I don’t think they would imagine some contractor somewhere is analyzing the conversations."