Together AI experienced a security incident in June and July 2026 where an unauthorized party accessed stored prompt samples. This occurred after a credential for a private Hugging Face repository was accidentally exposed in a public GitHub repository by an AI coding agent. The exposed prompts were from requests made to the Minimax M3, GLM-5.2, and Kimi K2.6 models. Together AI is investigating the extent of the data exposure and has implemented enhanced security measures, including scoped credentials and zero data retention as a default setting. AI
IMPACT Highlights the risks of prompt data exposure and the need for robust security in AI inference services.
RANK_REASON Security incident report from an AI service provider.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →