Together AI says an unauthorized party downloaded a subset of prompts submitted to its serverless inference service in June and July 2026. The company is reviewing the samples to determine what customer information was exposed and whom it needs to notify directly.
The samples came from requests to the MiniMax M3, GLM-5.2 and Kimi K2.6 models, primarily on June 24, July 4, July 8 and July 14, 2026. Together AI says the samples were authorized for routine software testing and stored in a private Hugging Face repository.
An AI coding agent exposed the repository credential
According to the company, an AI coding agent committed a credential for that private repository while working on Together AI’s public xorl GitHub repository. Automated scanning detected the exposed credential, and the company reverted the commit. During the exposure window, an unauthorized party obtained the credential and used it to download the samples.
Together AI says the party searched the samples for credentials to other services and tested whether those credentials worked. Five of the credentials tested were canary credentials associated with a Dragos security experiment. The company says it has no reason to believe the prompt data was leaked further and understands that it has since been deleted.

Together AI strengthened its security controls
Together AI says it has worked with an external forensic investigation firm and law enforcement. Its review has found no evidence of additional prompt leakage beyond this incident. The company also says it strengthened controls around development workflows, credentials and customer data. It advises customers handling sensitive prompt data to enable ZDR in their account settings.
The review of affected prompts is ongoing. Together AI says it will contact affected customers directly as needed with findings relevant to them.



