HiddenLayer launches runtime security product for AI coding agents
HiddenLayer said the new product extends its AI Security Platform to monitor and stop risky actions by AI coding agents inside developer workflows.
HiddenLayer said on Aug. 3 that it launched Agent Harness Security, a runtime security product for AI coding agents, in Austin, Texas.
The company said the tool is designed to help security teams protect source code, secrets and infrastructure from unsafe agent actions. HiddenLayer said the product extends its AI Security Platform’s Runtime Security module and is available now as a stand-alone solution.
HiddenLayer said the system plugs into each coding agent’s native hook surface to detect and stop prompt injection, sensitive data exposure and unsafe command execution as they happen. The company said it also provides visibility across prompts, tool calls, shell commands, file edits and repository interactions.
The announcement comes as AI coding agents become more common in software development. HiddenLayer cited Gartner as predicting that by 2028, 90% of enterprise software engineers will use AI code assistants, up from less than 14% in early 2024. The company said these agents can run shell commands, modify repositories, install dependencies and act on context from files, tool outputs and external sources.
Chris Sestito, HiddenLayer’s CEO, said securing AI coding agents requires more than allowing or blocking actions.
“Teams need to control what agents see, reason over, and act on before something goes wrong,” Sestito said in the release.
The release also included a quote attributed to Jaclyn Miller, CPO, and Vanessa DeGennaro, VP of Engineering, who said visibility, control, monitoring, governance and constraint are important as organizations scale AI coding agents across engineering.
HiddenLayer said the product can redact sensitive information before a model sees it, steer agents away from poisoned tool responses and stop unsafe actions before they pose a risk. The company said interested users can visit its coding-agents webpage or email the contact address provided in the release to schedule a demo.
Sources
Editorial note: This article was prepared with AI assistance and includes source attribution for reader review.
Comments ()