HiddenLayer

HiddenLayer builds security software for enterprises and governments deploying predictive, generative and agentic AI. Its AI Security Platform helps security teams inventory AI assets, scan models for supply-chain risks, test defenses through adversarial attack simulation and monitor applications at runtime. Developers can integrate these controls into CI/CD and MLOps workflows. Its Agent Harness Security extends runtime controls into AI coding agents through native hooks, allowing teams to redact secrets, detect prompt injection and block unsafe commands within each agent platform’s enforcement capabilities.

Founded in 2022 by Christopher Sestito, Tanner Burns and James Ballard, HiddenLayer grew out of experience responding to an attack against a machine-learning model at Cylance. Sestito remains CEO; Burns is chief scientist and Ballard is CIO. Its research examines how attackers can compromise models themselves. ShadowLogic demonstrated how backdoors embedded in a neural network’s computational graph can alter model behavior without conventional malicious-code execution and persist through fine-tuning, exposing risks beyond poisoned training data.

In 2024, HiddenLayer scanned models across Microsoft’s Azure AI model catalog, which contained nearly 1,700 models. The company raised a $50 million Series A in 2023, led by M12 and Moore Strategic Ventures, and reported protecting Fortune 100 companies’ AI models across finance, government and defense, and cybersecurity.