Contractors hired to train Meta's AI say they read intimate conversations between users and the Meta AI chatbot — and the data they review can identify the users behind those messages, Business Insider reports.

The report lands as Meta pushes its AI assistant across WhatsApp, Instagram, Facebook and Messenger, positioning the chatbot as a everyday companion millions of people talk to daily. Behind that product sits a human supply chain: contractors who label, rate and review conversation data so the model improves. According to the contractors interviewed by Business Insider, the job routinely exposes them to the private side of those exchanges.

What do the contractors actually see?

The workers say the material includes intimate talks users have with the chatbot — conversations many users likely assume stay between them and the machine. That assumption does not hold on the training side, according to the report.

The more serious disclosure concerns identity. The contractors say the data they handle is not fully anonymized, and includes information that can identify the users behind the chats. In practice, that means a labeler reviewing a conversation may be able to connect it to a real person.

Business Insider's reporting is based on accounts from the contractors themselves. The workers described reviewing exchanges that users plausibly treated as confidential, given the personal, sometimes intimate nature of chatting with an AI companion.

Why does this matter for Meta?

Meta has marketed its AI assistant as a conversational product people can talk to freely. If users learn that human contractors read those conversations — and can potentially identify who said what — that pitch takes a direct hit.

The disclosure also feeds a broader debate over how AI companies handle training data. Large language models need human review to improve, and human review means human eyes on real conversations. The question is what safeguards separate the user from the reviewer. According to the contractors' account, that separation is weaker than users might expect.

For regulators, the detail that identifying information reaches contractors is the sharpest point. Anonymization failures in AI training pipelines have already drawn scrutiny in the EU under GDPR and in the US under various state privacy laws. A report suggesting identifiable user data flows to third-party contractors will add pressure.

What has Meta said?

Business Insider's report centers on the contractors' accounts of their work. The publication contacted Meta for the story; readers should consult the original report for the company's full response.

Neither Meta nor its staffing partners have, to date, detailed publicly which fields of user data reach training contractors, how identifying information is stripped — or whether it is stripped at all before human review. That gap between the product promise and the pipeline reality is exactly what the contractors' accounts expose.

The bottom line

Users of Meta AI now have a reported answer to a question many never thought to ask: who reads your chats with the chatbot? According to the people doing the reading, the answer includes outside contractors — and they can see enough to know who you are.