$30m · Series A · AI · Boston, US and London, UK
Boston- and London-headquartered Mindgard, which tests and protects enterprise AI systems, has raised $30m in Series A financing. Album VC led the round, with Karma Ventures and existing investors .406 Ventures, Atlantic Bridge, IQ Capital and Lakestar participating. The company says it will expand product, engineering, sales and marketing.
AI security becomes a continuous systems test
Mindgard sells a security platform for the models, agents, tools, data and workflows that make up an AI application. Its discovery product maps deployed components and shadow AI; its red-teaming platform profiles a system, builds attack paths, executes adaptive tests and produces evidence for remediation. Runtime monitoring then applies the resulting intelligence to production behaviour.
The commercial mechanism is repetition. A model update, new prompt, changed permission or added integration can create a fresh attack path, so a one-off assessment ages quickly. Mindgard can turn that churn into recurring demand by fitting repeated tests into CI/CD, security operations and ticketing workflows. The value is not simply finding unusual model responses; it is giving engineering and security teams a reproducible path from exploit to fix and retest.
Research credibility has to become enterprise evidence
Mindgard grew from more than a decade of AI-security research at Lancaster University. In its funding announcement, the company said its platform had contributed to more than 150 publicly disclosed security and safety findings across widely used AI products. It also said the platform was deployed by Fortune 2000 companies and AI builders across financial services, pharmaceuticals, gaming, digital services, semiconductors and healthcare, although it did not name customers or disclose contract metrics.
That makes evidence quality central to the scale-up. Album VC's rationale is that enterprises are moving AI into critical operations without security infrastructure designed for model behaviour and connected tools. Mindgard's opportunity is to make attacker-style testing a standard control rather than a specialist project. Its execution test is whether automated findings remain reproducible across changing systems and translate into fixes that customers can verify, audit and buy repeatedly.
