Sign in Subscribe
Western Hemisphere

Mindgard Raises $30 Million to Protect AI Systems

The Editor

The Editor

— 4 min read
Share
Mingard CEO James Brear and Founder Dr. Peter Garraghan
Mingard CEO James Brear and Founder Dr. Peter Garraghan

Mindgard has raised US$30 million in a Series A round led by Album VC, with participation from Karma Ventures, .406 Ventures, Atlantic Bridge, IQ Capital, and Lakestar, bringing total funding to roughly US$42 million. The company operates from London and Boston and sells an automated red-teaming and runtime protection platform for artificial intelligence systems.

Mindgard was founded in 2022 by Dr. Peter Garraghan as a spinout from Lancaster University, drawing on research from what the company describes as the largest AI security laboratory of its kind; Garraghan now serves as chief science officer. James Brear, previously chief executive of Swimlane, Veriflow, and Procera, became CEO in October 2025. The platform performs three functions: discovering shadow AI that organizations do not know they are running, red-teaming deployed models and agents to find vulnerabilities, and monitoring them in production. Its research team has publicly disclosed more than 150 high-impact vulnerabilities in widely used AI products, including a zero-day in the Cursor development environment that permitted arbitrary code execution without phishing or malware, weaknesses in ChatGPT's image-generation safeguards, and data exposure across trust boundaries in Google Antigravity. Customers span Fortune 2000 companies in financial services, pharmaceuticals, gaming, digital services, semiconductors, and healthcare. "AI is creating an entirely new attack surface and organizations need a fundamentally different approach to securing it," Brear said. The funding will scale product, engineering, sales, and marketing.

Market Context

The category Mindgard occupies is expanding quickly and consolidating at the same time. AI security platforms were valued at roughly US$15.8 billion in 2026, with projections near US$56.5 billion by 2033. Two of the more visible independent players have already been absorbed: Palo Alto Networks acquired Protect AI, and Check Point acquired Lakera in September 2025, which means the largest security vendors are buying rather than building capability in this area.

The demand driver is the gap between AI deployment and AI governance. Organizations are moving models and agents into production faster than security teams can assess them, a mismatch Album VC partner Ty Boswell described as organizations "moving AI into critical operations without security infrastructure designed for how these systems operate in practice." Mindgard's disclosures illustrate the point: the vulnerabilities it found were not in experimental research systems but in Cursor, ChatGPT, and Google Antigravity, products used daily by developers and enterprises. Traditional application security tools were not built to test model behavior, prompt handling, or agent permissions, which is the opening the company is selling into.

The Signal

"AI is creating an entirely new attack surface and organizations need a fundamentally different approach to securing it." — James Brear, CEO, Mindgard

Regional Relevance

For the United States: Mindgard maintains a Boston office alongside its London base, and its customer profile, Fortune 2000 firms in financial services, healthcare, and semiconductors, points at the American enterprise market as the commercial priority. The consolidation pattern reinforces that: both Protect AI and Lakera were acquired by security vendors with US operations, which means a well-funded independent in this category is also a plausible acquisition target for the same buyers. For US enterprises deploying AI agents with access to internal systems, the vulnerabilities Mindgard has published in mainstream developer tools are a concrete argument that existing application security coverage is incomplete.

For the United Kingdom and European deep tech: Mindgard is a university spinout that has now raised roughly US$42 million while keeping its research roots in Lancaster, a route that British academic institutions have struggled to reproduce consistently despite strong research output. The involvement of European funds including IQ Capital, Lakestar, Karma Ventures, and Atlantic Bridge alongside an American lead investor reflects the typical structure for European deep tech at this stage. For the UK specifically, retaining a company that discovered vulnerabilities in products from OpenAI and Google is a meaningful data point in an ongoing debate about whether British research commercializes at home or relocates.

The Other Side

Does publishing vulnerabilities in major AI products create commercial tension? Mindgard's credibility rests substantially on disclosures affecting Cursor, ChatGPT, and Google Antigravity, which is effective marketing but positions the company adversarially toward the model providers whose products its customers use. Responsible disclosure is standard practice in security, though the dynamic differs from traditional software, where researchers rarely depend on the vendors they expose for the continued relevance of their own product.

Is an independent AI security company a durable business or an acquisition? Protect AI went to Palo Alto Networks and Lakera to Check Point, and the largest platform vendors have both the distribution and the incentive to fold AI red-teaming into suites enterprises already buy. A US$30 million round funds expansion, but the category's recent history suggests the realistic outcome for well-executed independents is a sale rather than standalone scale.

Can automated red-teaming keep pace with the systems it tests? Mindgard's approach is to operationalize expertise into automated testing, but models, agent frameworks, and tool integrations change on a timescale of weeks, and each new capability creates attack surface that did not exist when the tests were written. The company has not published data on detection rates, false positives, or how quickly its coverage adapts to new model releases.

Sources & Transparency

Read more