Braintrust, an AI evaluation and agent-infrastructure startup valued at $800 million, disclosed unauthorized access to one of its Amazon Web Services cloud accounts. The compromised account held customer API keys used to access cloud-based AI models from providers such as OpenAI and Anthropic. Braintrust confirmed one impacted customer, said it found no evidence of broader exposure, and asked all customers to rotate their stored API keys as a precaution while it investigated the root cause.