Mindgard

Mindgard is an AI security platform from a company spun out of Lancaster University AI security research, headquartered in Boston and London. It performs attacker-style reconnaissance and adaptive red teaming against AI models, agents, and applications, then provides evidence and remediation guidance. The company describes a continuous discover, recon, attack, and defend workflow.

What It Is

Mindgard is an offensive security platform built for AI systems and agents. It covers AI applications, chatbots, agentic workflows, APIs, connected tools, and supporting infrastructure, including RAG applications and systems connected through MCP or A2A servers. It is aimed at application security, product security, red team, AI/ML engineering, and governance, risk, and compliance teams.

How the Discover, Recon, Attack, Defend Flow Works

• Discover: AI agent evaluation and security scanning, AI-BOM and shadow AI risk exposure, and automated AI infrastructure crawling.
• Recon: AI attack surface enumeration, psychometric agent profiling, and agent profile and guardrail busting that map models, agents, tools, and behaviors before attacks run.
• Attack: AI red teaming, agent security testing, and AI risk compliance reporting that validate which weaknesses are exploitable.
• Defend: runtime AI protection and response, automated agent hardening, and defense audit.

Teams can retest fixes and verify defenses as models, prompts, tools, and policies change.

Deployment and Integration

Mindgard says it can be operational in minutes through APIs, CI/CD pipelines, Burp Suite, or a single-click workflow. The site lists integrations spanning open-source and managed AI platforms.

Research Basis

Mindgard states its research has produced 150+ publicly disclosed AI vulnerabilities, including findings in Google Antigravity, OpenAI Sora, Zed IDE, and xAI Grok, and that this intelligence feeds its proprietary knowledge base. The company also lists SOC 2 Type 2 compliance.