Trustwise

• Trustwise has put three different product labels in front of buyers since 2024, and no cited source says whether they are one product renamed or successive offerings, so a buyer who evaluated it last year cannot tell what relation that has to what is sold now.
• No developer documentation, interface specification or code appears among the sources reviewed, which leaves a technical evaluator only marketing-level description before a sales conversation.
• No SOC 2, ISO 27001 or other audit report of Trustwise appears on the probed surfaces, while the organizations it names are a health service and two banks, so a buyer of that kind has nothing from an outside auditor to weigh.
• The named organizations appear as testimonials, a partnership quote and one proof of concept, so nothing in the record establishes a paying customer.
• Trustwise markets the control as model-agnostic and installable without replacing anything, so what a departing customer would have to rebuild is the policy library it compiled into Trustwise's own runtime checks, which no cited source sizes.

Enterprises are handing AI agents authority over live systems, and the actions land before anyone reviews them. Trustwise puts it in operational terms on its homepage, that agents call APIs, move money, change records and trigger downstream systems, and that a single intention can fan out into many side effects an enterprise sees only after the fact.

The buyer it addresses is whoever carries that exposure inside a regulated organization. Its customer testimonials come from healthcare, financial services and life sciences, which it calls three of the most heavily regulated industries there are.

No cited source puts a number on the problem. Nothing in the reviewed record counts how often agent actions go wrong, what they cost, or how many enterprises have hit the failure Trustwise describes, so the scale of the pain rests on the company's own telling. [s1, s2]

The control sits between the agent and the systems it acts on. Trustwise states that every agent action is intercepted before execution, evaluated against the customer's policies, and then authorized, blocked, or routed for human review, and its product page repeats that as authorizing every action before it runs and verifying every decision against policy, scope and authority.

The platform is sold as three modules that feed each other. Assess maps what each agent can reach and what authority it holds and classifies risk before deployment, Control authorizes and verifies each action at runtime, and Optimize uses the resulting record to widen what agents are permitted to do. The compliance page describes customer policy as something the platform compiles into runtime checks, evidence requirements and escalation rules, beside a mapped library the company puts at more than 1,100 controls drawn from the Responsible AI Institute and aligned to OWASP, ISO, NIST, HIPAA and the EU AI Act.

The detail stops where the marketing stops. The reviewed record carries no developer documentation, no interface specification and no code, so an evaluator cannot check how interception works, what it costs in latency on their own traffic, or how a policy is written. The performance and coverage figures on the site are the company's own, and no cited source tests them. [s1, s2, s3, s4]

Trustwise positions against categories rather than named rivals. Its homepage sets prompt filters that inspect words, observability tools that report after the fact, and cloud security that protects infrastructure against its own claim to control agent behavior before execution.

The one comparison in the record comes from Trustwise itself. A Grit Daily piece likens Trust Posture Management to the way Wiz defined cloud security posture management, and argues that model vendors such as OpenAI, Anthropic and Google build safety mechanisms around their own models while enterprise AI runs many models at once. Its press index records a listing in Gartner's 2025 Cool Vendors for agentic AI in banking and investment services, and the reviewed record carries the company's announcement of that listing rather than the report itself.

The same positioning cuts back. Trustwise markets the control as model-agnostic, cloud-agnostic and installable without rip-and-replace, and what a departing customer would carry is the policy library compiled into Trustwise's own runtime checks, which no cited source sizes. [s1, s2, s6, s11]

Almost every named organization reaches the record through Trustwise. Its product page carries testimonials attributed to officers at NHS, at Lloyds Banking Group and at Lyric Health, and an account on the company's own site describes work with NatWest Group under the Harmony AI Control Tower name, which that text calls a proof of concept. The exception is Health Innovation Kent Surrey Sussex, whose chief executive is quoted in Help Net Security describing a partnership to deploy agentic AI with a trust layer.

The commercial machinery is defined. Packaging is published as three tiers, one module against a single prioritized use case, all three modules across a defined set of use cases, and an enterprise tier covering unlimited use cases, the company says it prices by module, platform scope and deployment rather than by agent count, and no price appears on the page. The press index records a chief revenue officer appointed in April 2026 and a partner-program membership with HPE announced in July 2026.

What is missing is evidence of a purchase. No cited source names an organization that bought the product, the production claims in the record are the company's own words carried by others, and no revenue, customer count or renewal figure appears in the reviewed record. [s2, s5, s6, s10, s11, s13]

The founder's record is the part an outsider can check. theCUBE Research, an analyst outlet writing in June 2024, states that Manoj Saxena ran IBM Watson as its inaugural general manager, founded the Responsible AI Institute, has built and sold four companies and holds 34 software patents in AI and web services, with Trustwise his sixth startup. A Grit Daily piece updated in August 2026 names him founder and chief executive.

Below him the record thins quickly. Help Net Security named Matthew Barker as head of AI research in June 2025, and Trustwise's press index records Gina Carfagno's appointment as chief revenue officer in April 2026. No other executive, board composition or headcount figure appears in any cited source.

No team page appears among the pages reviewed, so beyond those three the people building the product are not publicly identified. [s6, s9, s10, s11]

No audit report appears in the reviewed record. A probe of trust.trustwise.ai returns a marketing page offering an AI agent readiness test rather than a trust center, and no SOC 2, ISO 27001 or comparable attestation appears on the probed surfaces.

What the company offers instead is framework mapping for the customer's AI, not for itself. Its compliance page states that it turns regulatory obligations into runtime controls mapped to NIST AI RMF, OWASP, HIPAA, the EU AI Act, ISO/IEC 42001, MITRE ATLAS and ISO/IEC 21031, and that customers can author their own policies for the same engine to enforce. Mapping a customer's obligations is a product feature rather than evidence about Trustwise's own security.

That gap sits awkwardly beside the organizations it names. Trustwise asks a bank or a health service to put its software in the path of their AI traffic, and the reviewed record carries no audit of Trustwise's own security to set against that. [s2, s4, s12]

Add analyzed competitors to compare them side by side with Trustwise.

Trustwise aims at regulated sectors rather than at a size of company. The organizations it quotes come from healthcare, financial services and life sciences, which it calls three of the most heavily regulated industries there are.

The named organizations match that framing. Officers at NHS, Lloyds Banking Group and Lyric Health appear in testimonials on the product page, and an account on the company's site describes a proof of concept with NatWest Group. Each is the kind of organization whose AI deployments already sit under an external supervisor.

What the segmentation lacks is a purchase. Nothing in the cited record establishes that any of those organizations bought the product, so the segments describe intent, a set of quotes and one proof of concept rather than a measured customer base.

The claimed advantage is where the check happens. Trustwise states that every agent action is intercepted before execution, evaluated against the customer's policies, and then authorized, blocked, or routed for human review, and it contrasts that with prompt filters that inspect words and observability tools that report after the fact.

Around that check sits a control library. The compliance page describes customer policy compiled into runtime checks, evidence requirements and escalation rules, with more than 1,100 controls drawn from the Responsible AI Institute and aligned to OWASP, ISO, NIST, HIPAA and the EU AI Act, so a customer's own rules run through the same machinery as the mapped regulatory ones.

The engineering behind it is asserted rather than shown. Trustwise says it turns autonomous AI into controlled enterprise infrastructure using proprietary evaluators, guardian agents, mapped controls and runtime evidence, and Help Net Security described the engine behind Harmony AI as combining red-teaming datasets, tuned small language models, synthetic personas and mapped controls. No documentation, benchmark or code that would let a buyer check any of that appears in the reviewed record.

Every path to the product in the reviewed record runs through the company. Packaging is published as scope tiers with no rate attached, no self-serve path appears among the pages reviewed, and Trustwise hired a chief revenue officer in April 2026 to lead enterprise growth and go-to-market expansion.

One channel is visible beside that. The press index records Trustwise joining the HPE Unleash AI partner program in July 2026 to bring runtime control of agentic AI to enterprise customers, which puts the product alongside a hardware vendor's AI stack.

The proof mostly stays inside the company's own pages. NHS, Lloyds Banking Group, Lyric Health and NatWest Group all reach the record through Trustwise, as testimonials or as a proof of concept, and the one partner named outside them is Health Innovation Kent Surrey Sussex, whose chief executive is quoted in Help Net Security. No cited source reports a purchase, a contract value or a renewal.

Trustwise publishes a shape but not a number. Three tiers are described, a single module against one prioritized use case, all three modules across a defined set of enterprise use cases with shared policy and evidence, and an enterprise tier covering unlimited use cases and surfaces.

The metric is deliberate. The company says it prices by module, platform scope and enterprise deployment rather than by agent, on the argument that one enterprise use case can involve dozens of agents, which keeps the bill from moving every time a team adds another agent.

No price, no minimum and no unit rate appears anywhere in the reviewed record, so the tier a buyer lands in and what it costs are both settled in the sales conversation.

The product is delivered as a layer the customer's existing stack routes through. Trustwise describes it as sitting between AI workloads and the systems that secure, observe and operate them.

It is sold as portable across the stack a customer already runs. The company says it is model-agnostic, cloud-agnostic and framework-compatible, that it integrates with existing AI infrastructure so one runtime control model covers the whole stack, and that adopting it needs no rip-and-replace.

Operationally the record stops at the boundary. Nothing in the reviewed sources describes how the control is installed, what happens to an agent when Trustwise is unavailable, or where the evaluation runs.

No audit report appears in the reviewed record. A probe of trust.trustwise.ai returns a marketing page offering an AI agent readiness test rather than a trust center, and no SOC 2, ISO 27001 or comparable attestation appears on the probed surfaces.

The compliance the company documents is the customer's, not its own. Its compliance page turns regulatory obligations into runtime controls mapped to NIST AI RMF, OWASP, HIPAA, the EU AI Act, ISO/IEC 42001, MITRE ATLAS and ISO/IEC 21031, and lets customers author their own policies for the same engine to enforce.

For the organizations Trustwise names that distinction matters. It asks a bank or a health service to put its software in the path of their AI traffic, and the reviewed record carries no audit of Trustwise's own security to set against that.

Trustwise pitches itself as neutral ground. It says it runs across models, clouds, frameworks, tools and deployment environments and integrates with existing AI infrastructure so an enterprise can apply one runtime control model without locking its AI control strategy to a single vendor.

That neutrality is the ecosystem strategy, and the record shows nothing it locks in. The HPE partner-program membership is the one relationship the reviewed sources name, and none of them connects it to a hold on the customer, because the product is sold on fitting whatever stack the customer already runs.

The partner side of the ecosystem is thin in the record. The HPE partner-program membership announced in July 2026 is the one relationship the reviewed sources name, and no marketplace listing, reseller, systems integrator or technology alliance appears beside it.

One person carries the public record. theCUBE Research states that Manoj Saxena ran IBM Watson as its inaugural general manager, founded the Responsible AI Institute, has built and sold four companies and holds 34 software patents in AI and web services, with Trustwise his sixth startup, and A Grit Daily piece updated in August 2026 names him founder and chief executive.

Two other names appear. Help Net Security named Matthew Barker as head of AI research in June 2025, and the press index records Gina Carfagno's appointment as chief revenue officer in April 2026.

Beyond those three the reviewed record is empty. No director other than Saxena, whom the Form D lists as an executive officer and a director, no investor representative and no headcount figure appears in any cited source, and no team page appears among the pages reviewed, so a buyer evaluating who would build and support the product has three names to go on.