Smartflow vs. OneTrust: Request Path vs. Governance Program

OneTrust runs the AI governance program: inventory, risk assessment, policy, approvals and audit evidence, with runtime signals from the major AI platforms and AI Guard, a classifier applications call through a Python SDK or REST API. In September 2026 it announced CORIE and an AI Control Plane to enforce policy at the agent's tool call. Smartflow enforces policy inline on every model call and MCP tool call routed through it, on the customer's own infrastructure, and records each decision. They do different jobs, and a regulated firm can run both.

The short version

OneTrust is where an organization decides and documents what its AI may do: which systems exist, who owns them, what risk they carry, which policies apply and who signed off. It reaches into production through the AI platforms it connects to and through AI Guard.

Smartflow is where that decision is enforced on each call. Applications and agents send their model calls and MCP tool calls through it, on the customer's own infrastructure, and policy is applied before the call executes. When policy requires a person, the action pauses until an approver decides.

The question for a buyer is coverage: which agents and which calls each product can see and stop today.

What OneTrust announced in September 2026

In announcements on September 29 and 30, 2026, tied to its TrustWeek event, OneTrust introduced CORIE (Contextual Orchestration for Reasoning, Intelligence, and Evidence). OneTrust says CORIE "enforces those rules at the tool call," allowing an action, stopping it or flagging it for human review, and records the agent, request, purpose, outcome and governing policy. It also introduced an AI Control Plane, "an independent control layer between AI agents and enterprise systems" that "applies or orchestrates the policies, guardrails, and controls" deciding whether an action proceeds. An MCP Gateway that brings OneTrust into ChatGPT, Claude, Copilot and Glean is slated for private preview this fall.

OneTrust has not published an availability date or a deployment model for CORIE's enforcement or the AI Control Plane. Today it also lists a policy-as-code AI Policy Engine, MCP controls that "define and enforce which tools an agent is permitted to invoke" and Guardian Agents, without saying where they enforce. Its detailed documentation covers platform integrations and AI Guard, described below.

Feature Comparison

Where the decision is enforced

OneTrust's detailed documentation describes two runtime routes. The first is the platforms it connects to: it reads telemetry from Amazon Bedrock, Microsoft Foundry, Databricks and Google's agent platform, and lists guardrail enforcement through its Foundry integration. The second is code: applications that call AI Guard, through its Python SDK or REST API, can redact or block sensitive content before it reaches the model. AI Guard classifies content; its documentation does not describe authorizing tool calls.

Both routes depend on where an agent runs and how it was built. An agent on a self-hosted model, an inference server on premises, or a tool call into an internal system can sit outside both. Smartflow covers those calls because they are routed to it: the control sits on the path the call takes, whatever platform produced it. The trade is that a call has to be pointed at the gateway.

CORIE and the AI Control Plane may widen that reach. OneTrust says the Control Plane "applies or orchestrates" the controls that decide whether an action proceeds. Until the architecture is published, the coverage question can only be answered for the routes above.

Who decides, and what the record shows

Both products put a person in the loop. CORIE flags an action for human review, and governance teams handle exceptions in OneTrust's Governance Command Center. In Smartflow, an action that needs a person pauses as an approval ticket: pending until someone with approval rights approves or denies it, redeemed once when the action runs, and expired after 24 hours. With step-up, set per action, the approver re-verifies their identity before the approval counts, and the record shows who decided.

For an examiner, the useful question is what the record proves and where it lives. OneTrust's Evidence Ledger captures what was considered, which policy applied and what action was taken. Smartflow's records are hash-chained, carry RFC 3161 timestamps from a public timestamp authority or the customer's own, and are written to the customer's own SIEM, next to the rest of its security evidence. An optional write-once archive keeps them in Azure immutable blob storage in the customer's account.

Where OneTrust is strong

OneTrust is the more complete product for the governance program itself. It brings AI inventory, risk assessment, policy management and approval workflows together with privacy, consent and third-party risk on one platform, so an AI decision can draw on a firm's consent records, vendor assessments and existing policies. Its assessment templates cover the EU AI Act, NIST AI RMF and ISO 42001. OneTrust reports a Visionary placement in the inaugural 2026 Gartner Magic Quadrant for AI Governance Platforms, and by its own count more than half of the Fortune 500 use it, so a regulated firm's governance team often works in it already.

Its runtime approach also asks less of the network. Reading telemetry from Bedrock and Foundry, and enforcing guardrails through its Foundry integration, needs no change to how traffic flows. For a firm whose AI runs on those platforms, that is the shortest route to visibility.

For an organization deciding what its AI policy should be, assessing systems against it and proving the program to an auditor, OneTrust is a strong choice, and Smartflow does not attempt that job.

When to choose Smartflow

• Agents beyond the major platforms: self-hosted models, on-premises inference and tool calls into internal systems, governed on the same path
• Enforcement inside the perimeter: the control plane runs on your infrastructure, including air-gapped, so policy decisions and records stay inside your boundary
• A person decides, on the record: the action pauses, the approver re-verifies identity with step-up where policy requires it, and the record shows who decided
• The model path itself: routing, failover and caching on the same gateway that enforces policy

Running both

The two compose because they do different jobs. OneTrust keeps the inventory, the assessments, the policy and the approvals. Smartflow enforces that policy on each call and writes the runtime record to the customer's SIEM, where the governance program can draw on it. The cost is two systems whose policies have to stay aligned. The benefit is that the program and the enforcement point each do the job they were designed for.

Common questions

Does OneTrust enforce AI policy at runtime? Through guardrail enforcement in its Microsoft Foundry integration and through AI Guard, which applications call through a Python SDK or REST API. It also lists policy-as-code, MCP controls and Guardian Agents without documenting where they enforce. OneTrust announced tool-call enforcement through CORIE and an AI Control Plane in September 2026 and has not published an availability date.

Is OneTrust an AI gateway? OneTrust does not document a gateway or proxy on the model path. Its announced MCP Gateway connects AI assistants such as ChatGPT and Claude to OneTrust.

Where does OneTrust AI Guard run? Its classification service runs on a worker node in the customer's environment, and applications call it through a Python SDK or REST API. OneTrust's cloud defines classification behavior and receives summary metrics; prompts and responses stay local.

Does Smartflow replace OneTrust? No. Smartflow does not run AI risk assessments, approval workflows or privacy programs. It enforces policy on each call and records the result.

Can both be deployed together? Yes. OneTrust runs the program, and Smartflow enforces in the request path and writes its records to the customer's SIEM.

Related reading

The AI control plane · AI agent governance guide · Smartflow vs. Credo AI · Agent governance · Examination readiness · Model Context Protocol · Human-in-the-loop · AI sovereignty · EU AI Act · NIST AI RMF · Trust Fabric architecture · All comparisons

Verified against each vendor’s published documentation in October 2026. Vendor capabilities change. If something here is out of date, tell us and we will correct it.

OneTrust sources: onetrust.com, onetrust.com/solutions/ai-governance, onetrust.com/solutions/ai-governance/runtime-controls, developer.onetrust.com (AI Guard), and OneTrust’s September 29 and 30, 2026 announcements of CORIE and the OneTrust Platform.