Trustwise builds runtime control for enterprise AI agents. Its AI Control Plane checks every prompt, tool call, delegation and action an agent attempts against policy, scope and authority before it runs, then records evidence of each decision for security, risk, legal and audit teams.
The problem it addresses is that autonomous agents can act on enterprise systems faster than people can review them. Trustwise turns an organisation’s own policies and the major AI frameworks into checks that run at request time.
At A Glance
| Vendor | Trustwise |
|---|---|
| Category | AI safety, guardrails and runtime control |
| Identities covered | AI agents, their authority, access and actions |
| Deployment | Model, cloud and framework agnostic, including on-premises and edge |
| Agentic AI | Runtime authorisation of agent actions, behavioural contracts, allowable action graphs |
| Investor | Hitachi Ventures |
| Market classification | AI Safety & Guardrails; also Agentic AI Identity, AI Governance & Compliance |
What Trustwise Does
Trustwise starts from the view that agents should earn autonomy gradually: their risk is understood before deployment, their actions are bounded while they run, and the bounds are widened only when the evidence supports it. Policy is set centrally and evidence of each decision flows back to the teams accountable for it. Compliance is treated as something that happens while agents run, not a report written afterwards: customers bring their own policies and Trustwise turns them into runtime controls, alongside mapped requirements from NIST AI RMF, the OWASP LLM and Agentic Top 10, HIPAA, the EU AI Act, MITRE ATLAS, ISO/IEC 42001, ISO/IEC 21031:2024 and FCA Consumer Duty.
Products
Trustwise AI Control Plane
The AI Control Plane is the platform as a whole, used through an API, CLI, user interface, dashboard and reports. It runs low-latency evaluators for safety, policy and runtime control so that risk is detected without slowing agent workflows, uses domain-specific agents to enforce behaviour, and tests systems with adversarial prompts for injection, jailbreak, misuse and drift. It also routes and optimises AI calls to cut cost and supports energy-aware operations.
Assess
Assess comes before deployment. It classifies agentic systems by risk and autonomy, maps each agent’s authority, access and blast radius, and produces a control readiness report that shows gaps across policies and controls.
Control
Control sets and enforces the limits. A Behavioural Contract defines how an agent is expected to behave, and an Allowable Action Graph compiles the actions it may take. Controls are validated against red-team flight paths, scoped action authority is enforced at runtime, and a Control Effectiveness Proof shows the controls work.
Optimise
Optimise gives agents more room once they have proved reliable. It monitors drift, tunes Trust Scores and delivers a Trust Posture Improvement Loop, an Autonomy Expansion Plan and a Value Realization Report.
How It Works
The Control Plane is a central layer that sets the desired state across a fleet of agents and enforces it at request time. It sits between AI workloads and the systems that secure, observe and operate them, and plugs into existing AI infrastructure, including the LangGraph, Google ADK, OpenAI, AutoGen, Claude Agent SDK and CrewAI agent frameworks.
| Component | What it does |
|---|---|
| Policy layer | Enterprise policies and framework controls are compiled into runtime checks, evidence requirements and escalation rules. |
| Runtime enforcement | Authority, scope, risk and compliance are assessed for every request, and each agent action is authorised, blocked or escalated before it executes. |
| Evidence record | Every decision stores the actor, action, context, policy, decision and outcome. |
Identities And Systems Covered
| AI agents | An inventory of each agent’s authority, access and exposure |
|---|---|
| Agent actions | Prompts, tool calls, delegations and actions governed against scoped authority |
| Models | OpenAI, Claude, LLaMA, Mistral, Gemini and open-source models |
| Infrastructure | AWS, Azure, GCP, HPE, on-premises and edge |
Company Background
| Offices | Austin, New York and London. |
|---|---|
| Leadership | Manoj Saxena, founder and CEO; Gina Carfagno, Chief Revenue Officer. |
| Funding | Launched in July 2024 with a $4 million round from Hitachi Ventures. |
| Recognition | Included in the 2025 Gartner Cool Vendors for Agentic AI in Banking and Investment Services report; 2024 InfoWorld Technology of the Year Award in AI and Machine Learning: Development. |
Note: This overview is compiled from Trustwise’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.