Summary
- Mindgard has closed a $30 million Series A led by Album VC with participation from European and existing investors.
- The company grew from more than a decade of AI security research at Lancaster University and now operates from London and Boston.
- Funding will support product, engineering, sales, and marketing as enterprise AI systems move further into production.
AI security company Mindgard has raised $30 million in Series A funding as investment continues to move towards technology designed to test models, agents, and AI applications under adversarial conditions.
Mindgard said the round was led by Album VC, with participation from Karma Ventures and existing investors .406 Ventures, Atlantic Bridge, IQ Capital, and Lakestar. The company plans to use the financing across product development, engineering, sales, and marketing.
The business grew out of more than a decade of AI security research at Lancaster University and is now headquartered in London and Boston. Its platform combines automated discovery and adversarial testing with research into how AI models, agents, and applications behave when deliberately manipulated.
Mindgard says its technology has contributed to more than 150 publicly disclosed security and safety findings across AI products. The company lists work involving coding environments, model guardrails, and other AI applications among the research underpinning the platform.
The funding round arrives as enterprise AI security begins to separate into several distinct control problems. Conventional application-security tooling can still examine code, dependencies, APIs, and infrastructure around an AI service, but model behaviour introduces additional failure modes including prompt injection, unsafe tool use, data leakage, manipulated model outputs, excessive agent permissions, and weaknesses in safety controls.
Those risks become more consequential as models are connected to business systems rather than used only as isolated chat interfaces. An agent that can query company data, call external services, change records, execute code, or trigger workflows inherits privileges from the systems around it. Security assessment must therefore account for the model’s behaviour and the authority attached to it.
That has created a new procurement category around AI red teaming and runtime protection. Some products test models before deployment; others monitor interactions in production; some attempt to discover AI assets that have been deployed without central oversight. The market remains immature, and vendors use overlapping terminology for capabilities that are often technically different.
Mindgard describes its platform as covering shadow-AI discovery, AI red teaming, and runtime protection across models, agents, and applications. The company says adoption has expanded across financial services, pharmaceuticals, gaming, digital services, semiconductors, and healthcare, although those customer claims originate from the company rather than independently published usage data.
The Lancaster connection gives the business a notable UK research base even as the company expands internationally. Universities have become an important part of the European security ecosystem around artificial intelligence, particularly where commercial products depend on specialist adversarial research that is still changing more quickly than established testing standards.
Investment in the sector also reflects a shift in how AI risk is being budgeted. Early enterprise projects frequently treated AI security as an extension of data protection or conventional application security. Systems capable of autonomous action create additional requirements around identity, permissions, tool access, monitoring, model behaviour, and governance.
A larger funding round does not establish which technical approach will become standard. AI security remains fragmented, and enterprise buyers are still working out which controls should sit with application-security platforms, cloud providers, AI gateways, identity systems, model developers, or specialist testing companies.
Mindgard’s new capital gives it more resources to compete as those boundaries are being established. The commercial opportunity depends less on the number of organisations experimenting with AI than on how many move models and agents into environments where failures can affect real data, real permissions, and real business processes.
