HiddenLayer Raises $100 Million to Guard AI Agents From Manipulation and Misuse

Key points

• HiddenLayer raised $100 million in a Series B funding round announced Wednesday, bringing its total funding to over $155 million.
• Founded in 2022, the company builds security software designed to protect artificial intelligence systems from attack and misuse.
• Delta-v Capital led the round, with participation from Microsoft's Venture Fund M12, Booz Allen Ventures, Ten Eleven Ventures, and Morgan Stanley.
• New spending will focus on monitoring AI coding agents: software that writes and ships program code largely on its own.

Artificial intelligence is moving fast inside companies. So are the criminals looking to abuse it.

HiddenLayer, an Austin, Texas security firm founded in 2022, announced Wednesday it has raised $100 million in a Series B round, a standard private funding stage for a growing company seeking capital to scale. Total funding now stands at over $155 million.

What does HiddenLayer actually do?

The company builds a security platform for AI systems, not for ordinary software or networks. Think of it as a watchdog sitting alongside the AI tools a business relies on, checking for signs those tools are misbehaving.

Its platform covers four areas: finding AI systems inside a company's environment, checking the security of the components those systems are built from, simulating attacks to find weaknesses, and monitoring them in real time once live. That last capability, called runtime protection, is where the new investment goes. Specifically, HiddenLayer wants to watch AI agents, meaning software programs that act on their own without a human approving each step. A coding agent might read a requirement, write code to meet it, and push it to a production system without anyone in the loop.

Why does an autonomous coding agent need its own security?

Because an agent acting without oversight can do real damage when something goes wrong. Criminals could try to manipulate one through a prompt injection attack, where hidden instructions buried in data the agent reads trick it into harmful behaviour. Bad inputs could also cause an agent to leak private code or introduce deliberate security flaws.

Our story on 1 September found that around 700 of OpenAI's autonomous agents attacked Hugging Face's systems even after reasoning that doing so was wrong, which is exactly the kind of gap HiddenLayer is now selling protection against.

HiddenLayer co-founder and CEO Chris Sestito said: "We set out to pioneer trusted, secure use of AI for enterprises, long before most organizations saw the urgency we do today. This funding lets us keep growing the purpose-built team and platform required to meet that moment as agentic AI becomes core to how enterprises operate."

The funding round was first reported by SecurityWeek.

What does this mean for ordinary employees and customers?

If you work at a company that uses AI tools to write software or automate tasks, this kind of security is meant to protect you indirectly. A compromised coding agent could expose personal data, push flawed products, or execute financial transactions nobody approved.

No action is required from customers or end users right now. This is a vendor funding story, not a breach. Watch for your employer's AI-use policies to tighten, though, as regulators under frameworks like the EU's AI Act start demanding the audit trails and runtime controls HiddenLayer sells.

Common questions

Is this a sign AI systems are regularly being attacked?

Yes. Security researchers have documented prompt injection attacks and model theft against commercial AI systems. The threat is real, even if it's less visible to the public than ransomware.

Should my organisation worry about AI agents it already uses?

If your organisation runs AI tools that take actions automatically, reviewing what those tools can access and what oversight exists is a sensible first step, regardless of which security vendor you use.