Voice & Audio OpenAI Opens GPT-Live-1.1 to the API and Prices Silence at 2 Cents a Minute Coding Assistants Replit Pitted 32 Claude Agents Against NumPy Tuning: 18.9x Speedup While GPT-6.2 Swarms Collapsed at Scale Security GAO Covert Tests: All Four Top Chatbots Produced Viable Attack Plans in Under 12 Hours Once Guardrails Fell Meta Meta Puts Its 2030 Net-Zero Target Under Review as AI Data Centers Push Net Emissions Up 45 Percent Business Daniel Ek Is Backing Europa Fusion, a Stealth UK Startup Led by a Former DeepMind Researcher Betting on Pulsed Power Coding Assistants Stripe Tested 14 Coding Agents on 69 Payment Scenarios: 78% of Criteria Passed, but Only 29% of Runs Met Every Requirement Security A Transcription Firm Used by Police and Prosecutors Left Interview Transcripts and Body-Cam Footage on an Open S3 Bucket News UN Killer-Robot Talks End With a 2028 Deadline and No Binding Rules. About 100 States Wanted a Treaty; the US, Russia and Israel Blocked It Voice & Audio OpenAI Opens GPT-Live-1.1 to the API and Prices Silence at 2 Cents a Minute Coding Assistants Replit Pitted 32 Claude Agents Against NumPy Tuning: 18.9x Speedup While GPT-6.2 Swarms Collapsed at Scale Security GAO Covert Tests: All Four Top Chatbots Produced Viable Attack Plans in Under 12 Hours Once Guardrails Fell Meta Meta Puts Its 2030 Net-Zero Target Under Review as AI Data Centers Push Net Emissions Up 45 Percent Business Daniel Ek Is Backing Europa Fusion, a Stealth UK Startup Led by a Former DeepMind Researcher Betting on Pulsed Power Coding Assistants Stripe Tested 14 Coding Agents on 69 Payment Scenarios: 78% of Criteria Passed, but Only 29% of Runs Met Every Requirement Security A Transcription Firm Used by Police and Prosecutors Left Interview Transcripts and Body-Cam Footage on an Open S3 Bucket News UN Killer-Robot Talks End With a 2028 Deadline and No Binding Rules. About 100 States Wanted a Treaty; the US, Russia and Israel Blocked It

Hadrian Raises $40 Million to Stress-Test the New Attack Surface: AI Agents That Can Be Talked Into Leaking Data

Hadrian Raises $40 Million to Stress-Test the New Attack Surface: AI Agents That Can Be Talked Into Leaking Data

Dutch offensive-security startup Hadrian has raised $40 million in a round backed by Forgepoint Capital International and SmartFin, bringing total funding to roughly $55 million. The company says the money will push its Atlas and Nova platforms from external scanning into internal networks — and into the threat class growing fastest: AI agents manipulated into disclosing data they should not.

Hadrian, an Amsterdam-based offensive security company founded in 2021, has raised $40 million in a round backed by Forgepoint Capital International and SmartFin, bringing its total funding to roughly $55 million — a figure that includes an initial seed round of $11.73 million led by HV Capital in June 2022. The company employs about 105 people.

Its existing products, Atlas and Nova, use AI for continuous exposure testing and on-demand penetration testing — effectively running attacker logic against a customer's external footprint on a schedule, rather than once a year. The new capital funds expansion across EMEA and the US plus further engineering and research.

The strategic shift the money pays for is from outside the perimeter to inside it. Co-founder and CEO Rogier Fischer said the goal is to move beyond "we can get into system X" toward mapping what an attacker does next: "We can show you not just, 'Hey, we can get into any system X, Y, Z,' but rather, 'Hey, we can get into system X, and then we can propagate inside your network to that database that has all your customers in there.'" That view requires customer permission and, unlike external scanning, installing agents or routers inside the network — a genuine deployment change, not just a product checkbox.

The most interesting part of the thesis is qualitative rather than technical. Fischer argued the threat landscape is shifting away from purely technical vulnerabilities toward manipulating automated workflows, and described a case in which a customer used agentic workflows for customer support and the company was able to influence the agent into inadvertently sharing sensitive information about other customers. "It's really about deceiving our agents, deceiving technology," he said. Testing for that class of failure requires dynamic engagement with live endpoints rather than static scanning, which is precisely where an AI red team has an edge over a checklist.

The same agents are also being pointed at remediation. Hadrian plans to build agents that implement fixes — adjusting firewall configurations, correcting misconfigurations — rather than only reporting them, which matters most for third-party software customers cannot patch at the source. Fischer said the company expects to reach profitability in the coming year, and framed the round as a buffer rather than a war chest: "You don't want to over-raise because you might end up with a lot of cash in the bank that you're not going to use and dilute a lot more."

The wider read: penetration testing has long been a services business priced in consultant-days, and AI-driven continuous testing is turning it into software. As enterprises give agents real permissions over internal systems, "can your agent be socially engineered into handing over data" is becoming a board-level question — and, for vendors like Hadrian, a new product category to sell.

Comments (0)

Log in to join the discussion

Log In

No comments yet