Braintrust warns customers to rotate API keys after AWS breach

Braintrust told customers to replace stored API keys after confirming unauthorized access in one of its Amazon Web Services accounts, a breach that put a niche AI infrastructure vendor at the center of a larger supply-chain risk. The company said the affected environment held AI provider API keys that customers store with Braintrust, credentials that can let an attacker act as a legitimate user and reach connected cloud systems.

Braintrust’s trust center said suspicious behavior was reported in one of its AWS accounts on May 4, 2026, and that unauthorized access was confirmed. The company said it contained the incident, locked down the compromised account, audited and restricted access across related systems, and rotated internal secrets. It also asked every customer to rotate any API keys stored with Braintrust, a step that means replacing old credentials with new ones and disabling the exposed keys before they can be reused.

Braintrust told TechCrunch it had communicated with one impacted customer and had not found evidence of broader exposure. Spokesperson Martin Bergman said the warning to customers was sent “out of an abundance of caution.” The company said the cause of the breach remained under investigation.

The incident highlights how a breach at a relatively small AI infrastructure provider can ripple outward. Braintrust, which founder and CEO Ankur Goyal previously described as an “operating system for engineers building AI software,” raised $80 million in a Series B round in February at an $800 million valuation. TechCrunch said the startup’s platform is designed to monitor AI models and products, which makes it a useful hub for companies building and running AI services, but also a concentrated target for anyone hunting secrets.

Cybersecurity startup co-founder Jaime Blasco, who received Braintrust’s breach email alert, warned that the incident could have “downstream implications” for affected customers, including AI companies that rely on Braintrust. That warning fits a broader pattern in cloud security, where attackers increasingly aim at service providers and shared infrastructure to steal API keys, then move into downstream systems as if they belonged there. For AI vendors handling customer secrets, the bar is no longer just keeping attackers out of one account. It is proving that the security controls around every connected system are strong enough to stop one breach from becoming many.