Summary

• NeuralTrust has raised a $20 million seed round led by Alstin Capital, with backing from European and Spanish investors.
• The company builds security and governance tooling for enterprise AI agents spanning multiple platforms, endpoints, and workflows.
• The round reflects a shift from chatbot adoption to operational AI systems that need runtime controls, auditability, and shutdown mechanisms.

NeuralTrust has raised $20 million to expand its AI agent security platform, putting fresh capital behind one of the harder control problems in enterprise AI: how to govern autonomous systems once they begin operating across business workflows.

The seed round was led by Alstin Capital, with participation from VentureFriends, Seaya, Kibo Ventures, Banc Sabadell, EA Ventures Plug and Play Fund, and Finaves, the venture capital fund of IESE Business School. NeuralTrust is also backed by public funding from the European Innovation Council and Spain’s State Research Agency.

The company plans to use the funding for engineering, product integration, and European expansion. Its platform is designed to help organisations identify, secure, monitor, and govern AI agents as deployments move from experiments into production systems.

Enterprise AI adoption is changing shape. Many companies began with copilots, chat interfaces, retrieval tools, and productivity assistants. The next wave involves agents that can use tools, access applications, call APIs, and act on instructions across sales, customer service, finance, software development, operations, and compliance.

That shift changes the risk profile. A chatbot that gives a poor answer creates one kind of exposure. An agent with permissions, system access, and the ability to trigger downstream actions can expose sensitive data, call the wrong tool, bypass policy, or create an operational error that spreads beyond the original prompt.

NeuralTrust argues that enterprise agents rarely run inside one neat environment. They span platforms, endpoints, business units, and model providers. That can create fragmented security, particularly where teams adopt different AI tools without central visibility. The company positions its platform as an enforcement layer that helps organisations discover agents, apply controls, monitor behaviour, and respond to risky activity.

The company says customers include AirEuropa, Abanca, Iberia, and Banc Sabadell, alongside banks, airlines, energy companies, and government agencies. Most of its customers are large enterprises, according to NeuralTrust, and a large majority are based in Europe.

Those sectors explain why agent security is moving quickly from an innovation concern into operational risk. Banks, airlines, energy companies, and public agencies work under resilience, privacy, safety, and regulatory obligations. They cannot deploy autonomous systems as ordinary productivity software. Controls have to cover identity, access, data movement, logging, accountability, incident response, and escalation.

AI agents also challenge existing governance models. Security tools were designed around users, devices, applications, networks, and workloads. Agents can behave like a mixture of user, application, workflow, and automation script. They may act on behalf of employees, invoke APIs, generate content, modify records, and interact with external services.

The category is still forming, and buyers will need to separate useful controls from generic AI security branding. Platforms in this space will have to integrate with identity systems, cloud estates, application environments, model providers, policy engines, and security operations tools. Compliance teams will also need evidence that agent actions can be reconstructed, audited, and constrained.

European demand may be strong because AI governance is being shaped by the EU AI Act, data protection law, cyber resilience obligations, and sector specific rules. Agent security tools could become part of the infrastructure that allows organisations to adopt AI without losing control of permissions, accountability, and sensitive data.

NeuralTrust’s round shows that AI security is moving beyond model evaluation and prompt filtering. Once autonomous systems enter production, the essential test is whether organisations can see them, constrain them, investigate them, and stop them when their behaviour no longer matches the business process they were meant to support.