Via tech.eu
The Lancaster University spinout is betting that securing AI models requires a fundamentally different playbook than traditional cybersecurity
Mindgard, an AI security startup born from over a decade of research at Lancaster University, has raised $30 million to expand its platform for protecting AI systems against adversarial attacks, model manipulation, and data leakage.
The round marks a significant step up for the company, which had previously raised approximately $11.9 million in total funding through early 2025. Its most recent prior round was an $8 million raise closed in December 2024, led by .406 Ventures with participation from Atlantic Bridge, WillowTree Investments, IQ Capital, and Lakestar.
Why traditional security tools aren’t enough for AI #
Mindgard’s core product is an autonomous AI red teaming platform that simulates adversarial attacks, including jailbreaks and chained exploits, to find vulnerabilities before bad actors do.
Traditional application security tools were built for a world of deterministic software, where the same input always produces the same output. AI models don’t work that way. A generative AI system might behave perfectly in testing and then leak sensitive training data when prompted with a carefully crafted input in production.
The platform covers generative AI, multimodal systems, and agentic workflows. Mindgard’s technology allows enterprises to inventory their AI assets, assess vulnerabilities across models and applications, and establish controls against specific threat categories like model manipulation and data leakage in production environments.
From academic research to enterprise security #
The company’s roots trace back to Lancaster University, where its founding team spent more than a decade studying AI security before spinning out the commercial venture. Mindgard now operates from dual headquarters in Boston and London. The company has been building out its leadership team, adding a Head of Product and VP of Marketing in 2025.
The startup’s approach aligns with established security frameworks including MITRE ATLAS, which catalogs adversarial tactics against AI systems, and OWASP. Mindgard has also been featured in discussions around Gartner’s AI Trust, Risk, and Security Management framework, known as AI TRiSM. Industry analysts have categorized Mindgard as a specialized DAST-AI solution, essentially dynamic application security testing purpose-built for artificial intelligence.
The AI security market is heating up fast #
Adversaries can manipulate model outputs through carefully crafted prompts, extract proprietary training data, or exploit the chained tool calls that agentic AI systems rely on. These are documented attack patterns that existing firewalls and endpoint protection simply weren’t designed to catch.
The EU AI Act mandates risk management processes for high-risk AI systems, creating a compliance driver on top of the security imperative.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our