Mindgard — $30M Series A for AI security testing / red-teaming (August 2026)

Tag: S-2026-08-12-mindgard-series-a Type: article (independent trade press — FinTech Global — relaying the vendor’s funding announcement; fetched in full) Author(s): FinTech Global (jhaxell) Date of source: 2026-08-12 Date ingested: 2026-08-14 Authority weight: medium — independent outlet, but content closely tracks the vendor’s announcement and hedges capability claims (“what it describes as”, “what it says is”); funding facts stated as fact but not cross-checked against a second outlet. Raw file: /_raw_sources/S-2026-08-12-mindgard-series-a.md

What it claims

Mindgard (Boston/London; spin-out of 10+ years of research at Lancaster University) closed a $30M Series A led by Album VC, with Karma Ventures plus existing backers .406 Ventures, Atlantic Bridge, IQ Capital and Lakestar. The platform covers Shadow-AI discovery, AI red-teaming and runtime AI protection across “models, agents and applications”. The company says it has publicly disclosed 150+ significant security and safety flaws in widely used AI products (examples given: a zero-day code-execution bug in the Cursor IDE; a trusted-workspace weakness in Google Antigravity; ChatGPT image-generation guardrail shortcomings), feeding a proprietary knowledge base. It claims adoption “across a large share of the Fortune 2000” and in sectors including financial services, pharmaceuticals, gaming, digital services, semiconductors and healthcare. It also claims a newly identified “psycho-technical” attack surface within AI models, agents and applications, which its platform is “the first” able to capture and exploit. Funds go to product, engineering, sales and marketing.

Notable quotes

“We don’t just automate attacks. We operationalize expertise, turning the knowledge of leading AI security researchers and offensive security practitioners into the capabilities every enterprise needs to secure their AI.” — James Brear, CEO

What’s speculative vs. asserted

• Asserted by the outlet as fact: the round (amount, lead, participants), HQ locations, Lancaster University origin.
• Vendor-asserted, relayed with hedging: “world’s largest AI security lab”, “first platform”, the “psycho-technical attack surface” concept, Fortune 2000 adoption share, the 150+ disclosure count and individual disclosures.
• Sector listing only: “financial services” appears in a list; no named FS customer, no jurisdiction.

Topics this feeds

• AI Governance Platforms — capital re-entering the testing / red-teaming / evaluation segment (previously flagged as quiet on this page), from a UK-university-spin-out with an offensive-security framing; complements the pre-deployment-evaluation locus (Patronus) and the guardrail locus (Giskard, Lakera).

Open questions raised

• Does continuous adversarial red-teaming output map to evidence FS validators can use under EU AI Act Art. 15 accuracy/robustness, SS1/23 / SR 11-7 validation, or DORA threat-led testing — or is it security tooling whose findings never reach the model-risk file? [inference from this vault]
• Is the “psycho-technical attack surface” a substantive new category or branding on prompt-level manipulation? (Not determinable from this source.)