Mindgard raises €26M Series A to turn AI red-teaming expertise into enterprise security infrastructure

Mindgard (London) has raised €26 million / $30 million in a Series A led by Album VC, with new investor Karma Ventures and returning backers .406 Ventures, Atlantic Bridge, IQ Capital, and Lakestar participating. The company was founded in 2022 by Dr. Peter Garraghan and spun out of Lancaster University, where it remains anchored to what it describes as the world's largest AI security lab. Mindgard builds infrastructure for discovering, assessing, and defending enterprise AI systems — operating as an autonomous red team that maps and continuously tests the attack surface that AI models, agents, and applications create.

A new attack surface that existing tools weren't built for

Traditional application security assumes a relatively stable attack surface: code that does what it says, interfaces that respond predictably, data flows that can be audited statically. AI systems break every one of those assumptions. A large language model is not a deterministic function; an AI agent can be manipulated through the text it processes; an image-generation system can have its safety guardrails bypassed by inputs that never appear in training. Mindgard calls this the psycho-technical attack surface — the class of vulnerabilities that emerge specifically from how AI models represent knowledge and respond to prompts.

The company's platform addresses this through three functions: Shadow AI discovery (mapping what AI models and agents an organisation is actually running, including those deployed without formal IT approval), AI red-teaming (acting as an autonomous attacker to find exploitable weaknesses before adversaries do), and runtime AI protection (continuous monitoring and mitigation once systems are in production). Partners NCC Group and KPMG provide managed routes to market in regulated industries.

The concrete track record matters here. Mindgard has publicly disclosed more than 150 high-impact security vulnerabilities across widely deployed AI systems: a zero-day code-execution flaw in Cursor IDE, a trusted-workspace vulnerability in Google Antigravity, and image-generation guardrail failures in ChatGPT. These are production-system findings, not theoretical attack proofs.

Why the investor syndicate is a continuity signal

The Series A investor list reads as a confirmation round more than an introduction. IQ Capital, Lakestar, and Atlantic Bridge are all returning investors expanding existing positions — they have seen Mindgard's commercial traction first-hand and are choosing to put significantly more capital to work at the €26 million level. Album VC and Karma Ventures are new entrants who are joining a round priced above the typical early-Series A ceiling, a decision that implies they are arriving at Mindgard after scrutinising the customer base rather than the pitch deck.

"Organisations are moving AI into critical operations without security infrastructure designed for how these systems operate in practice," said Ty Boswell, Partner at Album VC. "Mindgard has translated deep research and elite offensive-security expertise into a continuously evolving capability that enables enterprises to stay ahead of emerging AI threats."

What the capital is for

The proceeds fund scale across product, engineering, sales, and marketing to address what the company describes as significant customer demand. Mindgard's current base includes Fortune 2000 companies in financial services, pharmaceuticals, gaming, digital services, semiconductors, and healthcare — sectors where AI is being deployed into operational workflows and where a failure traced to an undetected vulnerability would carry regulatory and reputational consequences far beyond the security team. The 18-month question is whether the platform can define an enterprise-grade standard for AI security fast enough to become the category reference before the market fragments.