Mindgard

Mindgard is an automated AI red-teaming platform that continuously tests LLMs, AI agents, and multimodal models for adversarial vulnerabilities like prompt injection and model compromise, built by a Lancaster University spinout.

Mindgard is a specialist AI security and governance platform focused on offensive security testing for AI systems. Rather than static checklists, it operates as an autonomous red teamer that maps, plans, and executes complex, agentic attack workflows against LLMs, AI agents, and multimodal models. Its techniques span reconnaissance, inference, evasion, insider-threat scenarios, prompt injection, code audit, and model compromise, surfacing how real adversaries could discover and exploit weaknesses in deployed AI. Testing is continuous, so security teams can monitor risk as models and prompts change. Mindgard grew out of Lancaster University as a spinout founded in 2022 and is based in London and Boston. It has demonstrated real-world impact by disclosing more than 150 vulnerabilities across popular AI products, including a zero-day code-execution flaw in a widely used AI coding IDE and defects in other major AI tools. The company raised a $30 million Series A to scale its platform and research, and appointed a new CEO in late 2025 as its founder moved to Chief Science Officer. Mindgard is best for security teams, AI platform teams, and enterprises deploying AI who need to test models and agents for adversarial risk on an ongoing basis. It is less relevant to organizations not yet shipping AI features, and, as a testing and red-teaming platform, it complements rather than replaces runtime guardrails and broader governance programs.