LatticeFlow AI provides a platform that turns AI governance requirements into technical tests. It is aimed at organisations that have to show regulators, auditors and their own risk teams that their AI systems are safe, secure and compliant.
The problem it addresses is evidence. Regulations and internal policies say what an AI system must do, but proving that a given chatbot, copilot, foundation model or agent meets them takes testing, and the result changes every time the system does.
At A Glance
| Vendor | LatticeFlow AI |
|---|---|
| Category | AI Model Security, AI governance and evaluation |
| Deployment | SaaS (fully managed) or on-premises and private cloud |
| Agentic AI | Evaluation of agentic workflows, tool access permissions, agentic CLI |
| Customers | Unique, Axpo, PastaHR |
| Certifications | SOC 2 Type 2 |
| Funding | Series A (2022) |
| Market classification | AI Model Security |
What LatticeFlow AI Does
AI governance frameworks set out requirements in prose, and a risk team still has to prove, system by system, that those requirements are met. LatticeFlow AI links each framework to specific risk controls and to evaluations that can be run against a live system. Every evaluation produces technical evidence with a record of where it came from, which can go straight to auditors or governance boards, and because the evaluations repeat as systems change, the evidence stays current rather than describing last quarter’s model.
Products
LatticeFlow AI Platform
The platform is built from five areas that follow an AI system from discovery to ongoing governance.
| Atlas | A public registry of more than 40 AI governance frameworks, each mapped to technical risk controls and ready-to-run evaluations |
|---|---|
| Discover | Continuous inventory of AI assets across the organisation |
| Evaluate | Evaluations of performance and security tailored to the use case |
| Secure | Security controls for AI systems, including adaptive red teaming |
| Govern | Governance across the lifecycle of each system, with continuous monitoring |
Controls come from frameworks including the EU AI Act, ISO 42001, NIST RMF, OWASP, MITRE ATLAS and FINMA. Risk scoring weighs the impact of a system, its domain, its deployment context and industry factors. Agents are evaluated as systems in their own right: the security controls around an agent, the permissions it has to call tools, and its behaviour under Agentic Probes, repeated whenever its prompts, tools or models change. Organisations without in-house capacity can have the same work run for them as a managed AI governance service.
COMPL-AI
COMPL-AI is an open source framework for assessing generative AI models against the EU AI Act, built with ETH Zurich and INSAIT on 27 benchmarks. It gives model developers and buyers a technical reading of what the Act requires of general-purpose models and a way to measure a model against it.
How It Works
The platform takes in observability traces, agents and retrieval augmented generation (RAG) systems, runs evaluations against controls from the Atlas registry, and stores the resulting evidence for governance and audit.
| Component | What it does |
|---|---|
| SaaS | LatticeFlow AI runs the platform, so there is no deployment or infrastructure work, and it connects to AI systems through their public endpoints. |
| On-premises or private cloud | The platform is delivered as virtual machine deployments managed through a control plane. It connects to private endpoints and data sources, and all data stays within the customer’s environment. |
| CLI and automations | A command line interface and automations plug into agentic workflows, agent skills and existing pipelines, so agents can trigger evaluations and monitoring themselves. The platform is also available through SAP Store for assessments across the SAP ecosystem. |
Company Background
| Origin | Founded by AI professors and researchers from ETH Zurich with expertise in formal methods, symbolic reasoning and machine learning |
|---|---|
| Headquarters | Zürich, Switzerland, with offices in San Francisco, Sofia and Dublin |
| Leadership | Dr Petar Tsankov (Co-founder and CEO), Dr Pavol Bielik (Co-founder and CTO), Prof. Dr Martin Vechev (Co-founder) and Prof. Dr Andreas Krause (Co-founder and Scientific Advisor), with Andre Boisvert as Chairman |
| Funding | A USD 12 million Series A in 2022, led by Atlantic Bridge and OpenOcean |
| Partnerships | SAP (2026) |
Note: This overview is compiled from LatticeFlow AI’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.