Meta Halts Collaboration with Mercor Amid Data Breach Investigation

Meta has indefinitely suspended its collaboration with data contracting firm Mercor to investigate a significant security breach that may have exposed sensitive AI training data. Other major AI laboratories are also reassessing their partnerships with Mercor in light of the incident.

Mercor, a key provider of bespoke datasets for AI labs such as OpenAI and Anthropic, confirmed the breach in a March 31 email to staff, stating that a recent security incident affected their systems along with thousands of other organizations worldwide. The breach is linked to a supply chain attack involving the open-source AI API tool LiteLLM, which was compromised by the hacker group TeamPCP. This attack potentially exposed companies and services that incorporated LiteLLM, including Mercor.

OpenAI has not halted its projects with Mercor but is investigating the incident to determine if proprietary training data was exposed. A spokesperson for OpenAI confirmed that the incident does not affect OpenAI user data. Anthropic has not immediately responded to requests for comment.

The breach has raised concerns about the security of proprietary AI training data, which is crucial for maintaining competitive advantages in the industry. The full extent of the data exposure and its potential impact on AI model development remain under investigation.