HiddenLayer announced a $100 million Series B on September 2 to expand software that helps enterprise security teams protect AI models and the agents built around them. The Austin company, founded in 2022, checks models before deployment and watches for attacks while AI applications are running. Its newest focus is coding agents that can change files and use software tools, not simply answer questions.
The product makes more sense when viewed through a developer's workflow. A coding agent may read a repository, retrieve outside material and issue a shell command to complete an assignment. HiddenLayer's Agent Harness Security is designed to trace those connected actions and flag attempts to redirect them. The 'harness' is the surrounding software that gives the model access to tools and files. HiddenLayer says it can block unsafe actions or redact sensitive information where the underlying agent platform supports enforcement; it is not claiming that every integration has identical controls.
Another part of the business starts before an AI model reaches users. HiddenLayer's supply-chain module inspects model files and their components for signs of tampering, and records their origins and versions. For a bank evaluating a third-party model for fraud detection or credit-risk work, that means checking what it is importing as well as monitoring the system after deployment. This is a use case described by the company's financial-services materials, not evidence of a particular bank's deployment.
The commercial customer is the organization adopting AI, rather than the consumer using its chatbot. HiddenLayer says it added more than 50 platform customers over the past year across sectors including banking, insurance, pharmaceuticals and government, while annual recurring revenue grew more than tenfold. Those are company-reported measures, not audited results established by Captables. The breadth of applications helps explain the product lineup: security teams need both an inventory of models entering the business and a way to investigate what deployed agents actually do.
Delta-v Capital led the Series B, joined by Ten Eleven Ventures, Morgan Stanley, Microsoft's M12 and Booz Allen Ventures. The company says the capital will support its agent-security products, hiring, channel relationships and international expansion, starting with Europe and the wider EMEA region. SiliconANGLE reported that the new round follows a $50 million Series A in 2023 and takes disclosed funding to about $156 million, including its seed round.
Captables' read: the business is moving from checking AI ingredients to supervising AI actions. The practical test is whether customers can catch consequential misuse without making development workflows too cumbersome. New capital provides room to build and distribute those controls; repeatable performance inside real customer environments will determine how useful they become.