|
Together.ai: Prompt Leakage for Requests
Together AI disclosed a security incident where an unauthorized party accessed stored prompt samples from their serverless inference service in June and July 2026 after a developer accidentally committed credentials to a public GitHub repository. The exposed prompts came from requests to three AI models, and the attacker searched them for additional credentials, though Together AI found no evidence of broader data exposure beyond this incident. The company has implemented security improvements including scoped credentials, pull request scanning, and zero data retention as default to prevent future incidents.
Read Full Article →
← Back to latest