HelmGuard lets customers set the risk level above which a human checks its risk decisions

Somebody has to put a number in the box. HelmGuard's customers can require a human check only above a risk level they set ^[11]. Jack Miller, the chief technology officer, told Tech Funding News that no competitor offers that ^[11]. Every assessment carries a confidence score, every conclusion links to a specific source, and a human reviews the result depending on how critical the decision is ^[10]. Below the line the customer drew, the agent's conclusion is the conclusion.

"Most compliance platforms were built to document a process, not to reach a conclusion," John Daley, the chief executive, said in the funding announcement ^[6]. He named the competition in his interview with Tech Funding News: Vanta, Drata and Secureframe create checklists for people to follow, he said, and HelmGuard's agents work through the checklists themselves ^[8]. "We have this phrase internally: decisions over documents," Daley told the outlet ^[9]. He spent eight years as an executive at Palantir, according to HelmGuard ^[3].

What runs today pulls risk, security and compliance data out of company documents and directly from source systems ^[4]. Specialised agents then handle third-party risk management, control gap assessments and checks on other AI agents ^[4]. The layer that evaluates how agents behave while they are running is what part of the seed money pays for ^[17]. Daley said the funding "lets us bring that capability to far more teams and extend it to governing the AI agents being deployed on other workflows" ^[7].

HelmGuard's case is speed. It says it assessed 1,250 counterparties for one US insurer in less than a week, and that its engineers moved that customer off its previous platform in under 10 days ^[13]. Across seven days, 1,250 assessments average about 179 a day ^[21]. HelmGuard did not name the insurer. A global telehealth and telecommunications customer cut first response times on customer assurance from days to minutes, the company said ^[14].

Callosum, a London AI scale-up, hired HelmGuard to design and run its security and compliance programme ^[15]. "We now have the capability of a mature security organisation and can compete at massive scale," Danyal Akarca, the company's co-founder and chief executive, said in the announcement ^[16].

The Verified Risk Network, also funded by the round, is meant to let companies exchange verified, current claims about risk instead of documents ^[18]. "An AI vendor's risk profile changes with every model update and every new tool its agents can call," Miller said in the announcement ^[19]. He said the network "makes the unit of assurance a claim assessed directly by an agent, rather than a document, and enables agent-to-agent exchange on a continuous basis" ^[20]. HelmGuard says its users sit in the US, Canada, the UK, Hong Kong and South Africa ^[12].

Ask what share of assessment volume falls under the threshold, because that share is where the vendor's model is the reviewer of record. Ask whether your auditor accepts the citation and reasoning trace HelmGuard shows as the answer to how a conclusion was reached ^[5]. A checklist tool leaves the judgement with your staff and bills you for the paperwork. An agent that reaches conclusions moves part of that judgement into a vendor's model, and the report still goes out under your company's name ^[5].