Mindgard, a startup focused on finding and fixing security vulnerabilities in artificial intelligence systems, has raised $30 million in Series A funding to expand its team and global reach. The London-based company, led by James Brear and Dr. Peter Garraghan, combines AI security research with offensive security expertise to identify weaknesses that attackers could exploit in AI models, agents, and applications.

The funding round was led by Album VC, with participation from Karma Ventures and existing backers including 406 Ventures, Atlantic Bridge, IQ Capital, and Lakestar. Mindgard plans to use the capital to scale product development, engineering, sales, and marketing in response to strong customer demand from Fortune 2000 companies across financial services, pharmaceuticals, gaming, digital services, semiconductors, and healthcare.

What Makes Mindgard's Security Approach Different?

Unlike traditional cybersecurity firms that focus on protecting networks and data, Mindgard addresses a fundamentally new problem: AI systems behave differently than conventional software, creating attack surfaces that existing security tools weren't designed to handle. The company has developed a platform that automatically identifies and exploits emerging vulnerabilities across AI systems before malicious actors can weaponize them.

The platform's track record speaks to the scale of the problem. Mindgard has uncovered and publicly disclosed more than 150 high-impact AI security and safety vulnerabilities, including flaws discovered in Cursor IDE, Google Antigravity, and ChatGPT. These findings continuously strengthen Mindgard's proprietary security knowledge base, enabling security teams to access evolving offensive and defensive capabilities for AI at scale.

Why Are Enterprises Suddenly Concerned About AI Security?

Organizations are deploying AI in critical operations, from financial decision-making to drug discovery, without security infrastructure designed for how these systems actually behave in production. This creates a dangerous gap: companies understand how to protect traditional software, but AI systems can fail in unexpected ways that traditional security testing doesn't catch.

The problem is urgent because AI models can be manipulated through prompt injection attacks, data poisoning, model extraction, and other novel techniques that don't apply to conventional applications. A vulnerability in an AI system used for fraud detection or medical diagnosis could have cascading consequences across entire organizations or industries.

How to Assess and Defend Your AI Systems Against Emerging Threats

• Conduct Offensive Security Testing: Use automated reconnaissance and intelligent penetration testing to identify vulnerabilities in your AI models before they're exploited by attackers in the wild.
• Establish a Security Knowledge Base: Build and maintain an evolving repository of known AI attack patterns and defensive strategies specific to your organization's AI deployments and use cases.
• Integrate Security Into AI Deployment: Make AI security a core part of your enterprise AI deployment and management process, not an afterthought added after systems go live.
• Monitor for Emerging Attack Surfaces: Continuously assess new threats as AI technology evolves, since attack vectors discovered today may not have existed six months ago.

Kristjan Laanemaa, Founding Partner at Karma Ventures, emphasized the importance of combining research with practical enterprise capabilities.

"AI security is an emerging category, and Mindgard stands out for combining world-class offensive security research with automated reconnaissance and intelligent penetration testing. The platform is already used by leading Fortune 2000 enterprises, combining a strong research foundation with proven enterprise demand," he stated.

Kristjan Laanemaa, Founding Partner, Karma Ventures

The Series A funding reflects broader recognition that AI security is no longer optional. As AI systems move from research labs into production environments handling sensitive business operations, the cost of a security breach grows exponentially. A vulnerability in an AI model could expose proprietary training data, enable model theft, or cause the system to make biased or incorrect decisions at scale.

Mindgard's approach differs from traditional security vendors because it treats AI systems as a distinct category requiring specialized knowledge. The company's platform doesn't just scan for known vulnerabilities; it actively attempts to break AI systems using techniques developed by its research team at Lancaster University, uncovering novel attack patterns before they become widespread.

The investment also signals that enterprise customers are willing to pay for specialized AI security tools. Mindgard's rapid adoption among Fortune 2000 companies suggests that organizations recognize the gap between their current security capabilities and the threats posed by AI systems. As AI becomes more central to business operations, this gap will only widen without dedicated security infrastructure designed specifically for these systems.