Published on: February 18, 2025 Updated 2 times since publishing
AI security is rapidly becoming one of the most pressing concerns in cybersecurity, with threats evolving as fast as the technology itself. Mindgard, a pioneering company in AI security testing, is tackling these challenges head-on with its industry-first Dynamic Application Security Testing for AI (DAST-AI) solution.
In this exclusive SafetyDetectives interview, Peter Garraghan, CEO and co-founder of Mindgard, shares insights into his journey from AI research to launching a groundbreaking security platform. As a Professor of Computer Science at Lancaster University and a fellow of the UK Engineering and Physical Sciences Research Council (EPSRC), Peter has dedicated his career to bridging the gap between academic research and real-world security solutions.
He discusses the most critical vulnerabilities facing AI systems today, the unique security challenges AI presents, and how Mindgard is shaping the future of AI security for enterprises and governments alike.
What inspired you to launch Mindgard, and how did your background in AI research influence its development?
As a Professor of Computer Science and a fellow of the UK Engineering and Physical Sciences Research Council (EPSRC), I have spent over a decade exploring how AI and machine learning systems reshape the world as both transformative technologies and emerging threats. Through my research at Lancaster University, I recognized that conventional application security methods were inadequate for addressing AI-specific risks. I assembled a team to develop the first commercial security tool for AI models. Mindgard translates this R&D into an enterprise-ready solution that safeguards against AI threats, merging cutting-edge expertise with real-world application.
Mindgard is the first Dynamic Application Security Testing for AI (DAST-AI) solution. Can you explain how this approach differs from traditional security testing methods?
It is a methodology used to identify vulnerabilities in a running application by simulating real-world attack scenarios, much like an external adversary probing for weak spots within the system. Unlike traditional security tools that analyze static code, DAST-AI tests an operational AI system, interfacing with LLMs, RAG, and all supporting components, to uncover vulnerabilities that only appear at runtime. Our solution integrates across the development lifecycle from model scanning to full-system testing before release. We’re especially proud that our DAST-AI solution was recently recognized as a recommended tool in the OWASP LLM and Generative AI Security Solutions Landscape Guide 2025.
AI vulnerabilities such as prompt injection and model theft are evolving rapidly. What are some of the most pressing security challenges facing AI deployments today?
Prompt injection, model theft, and data leakage are just the tip of the iceberg. Attackers can manipulate AI to extract sensitive information, bypass safety guardrails, or clone proprietary models at a fraction of the original cost. Unlike traditional software, AI continuously adapts, meaning its attack surface is constantly shifting.
While AI doesn’t necessarily create new types of cybercrime, it accelerates and scales existing threats while introducing new attack vectors. Take phishing, a long-standing cybersecurity risk: AI can generate highly sophisticated, personalized emails by pulling data from multiple sources, adapting in real time to bypass security filters. The same principle applies to deepfakes, automated reconnaissance, and generative AI-enabled fraud, all of which drastically lower the cost and effort of cyberattacks. That’s why we built Mindgard – to provide continuous, automated security testing that identifies and mitigates AI-specific threats before they’re exploited.
Since AI security is still an emerging field, how should companies balance AI innovation with the need for robust security measures?
Recent reports show that security and data privacy concerns are the main hindrances preventing enterprises from implementing AI in their operations. Consider a recent case: an international law firm blocked general access to several AI tools after detecting a “significant increase in usage” by its staff. Would these bans improve their work or make the firm more competitive? The answer is yes, although it introduces a new set of risks. AI systems must meet the same rigorous security standards as any other software. Enterprises should evaluate AI security based on four key criteria: autonomy control (defining guardrails and restricting tool access), attack surface management (discovering threats such as instruction injection and adversarial manipulation), adaptability (continuously evolving with real-time threat intelligence), and integration (seamlessly working with existing security frameworks). The best AI security solutions will balance automation with oversight, assessment through red teaming, and strengthening defenses without introducing new vulnerabilities.
As AI regulations and security frameworks begin to take shape globally, how do you see the role of AI security evolving in enterprise and government sectors?
Take, for example, the recent AI Action Summit in Paris, which called on public, private, and academic stakeholders to collaborate in building a trusted AI ecosystem. It is one of the signals that AI regulations and security frameworks are maturing. Enterprises and governments will be forced to shift from ad-hoc AI security measures to continuous, standardized risk assessment. Compliance with regulations like the EU AI Act and emerging global standards will require organizations to demonstrate security testing, governance, and auditability of their AI systems. This will drive increased adoption of automated AI security solutions, moving beyond traditional cybersecurity approaches to dynamic testing that ensures AI remains secure throughout its lifecycle.
What’s next for Mindgard? Are there any upcoming developments or research initiatives that you’re particularly excited about?
A key priority is expanding into the US market to strengthen our global presence and better support our growing customer base. We are also deepening our R&D efforts to tackle increasingly sophisticated AI threats, keeping our platform at the forefront of AI security. Additionally, we continue to enhance our platform’s capabilities, ensuring it evolves to meet emerging customer needs and regulatory demands.
About Mindgard
Mindgard is the leader in Artificial Intelligence Security Testing. Founded at Lancaster University and backed by cutting edge research, Mindgard enables organizations to secure their AI systems from new threats that traditional application security tools cannot address. Its industry-first, award-winning, Dynamic Application Security Testing for AI (DAST-AI) solution delivers continuous security testing and automated AI red teaming across the AI lifecycle, making AI security actionable and auditable. For more information, visit mindgard.ai
Peter Garraghan bio
Dr. Peter Garraghan is the CEO and co-founder of Mindgard, a Professor of Computer Science at Lancaster University, and an internationally recognized expert in AI security. Since 2014, he has supervised hundreds of AI and machine learning researchers, leads a major research lab, and oversees a significant portion of the UK’s PhD candidates specializing in AI security. He advises the UK government, contributes to the Alan Turing Institute, and plays a key role in DIST’s research reports on AI cybersecurity. With over 60 published scientific papers, Peter’s work bridges the gap between cutting-edge research and practical solutions, shaping the future of AI security through both academic leadership and industry innovation.