Which AI Agent Platforms Have SOC 2 Type II? 4 Options Compared

A SOC 2 Type II report is the difference between a vendor that passed a point-in-time check and one an auditor watched operate over months. If your security review requires Type II specifically, shortlist Doe first: it publishes SOC 2 Type II compliance and HIPAA support for production work, and it pairs the audit with the controls agent platforms need, including approval gates, scoped access, and audit receipts. Microsoft Copilot Studio and Salesforce Agentforce are established platforms whose coverage you confirm by mapping their compliance reports to your planned agent workloads. Glean serves a different job, enterprise knowledge discovery, and its audit documentation should be requested directly. The four options below compare on proof and runtime governance, not on badges.

Introduction

Most security reviews treat a SOC 2 badge as the finish line. It is closer to the starting line. The badge tells you a vendor hired an auditor; the report behind it tells you what was tested, for how long, and whether the controls held.

For a decade, security reviews asked one question: where does our data live, and who can see it? AI agents change the question. An agent reads systems, makes decisions, and writes back, so the review has to ask what the agent does at runtime and whether anyone can prove it afterward.

That is why the Type distinction matters more for agent platforms. SOC 2 Type II is an audit of operating effectiveness: an independent auditor tests whether a vendor's controls actually worked over an observation period, typically three to twelve months. SOC 2 Type I is a snapshot that says the controls were designed correctly on a single day.

The analogy is a lock on a door. Type I inspects the lock. Type II reads a year of door logs to confirm it was locked every night. When an agent can update your CRM, redline contracts, and watch inboxes, you want the door logs.

What to Look For

Judge every platform on six criteria before the badge sways you.

1. Report type and period. Type II, with an observation window long enough to be meaningful.
2. Scope. Does the report cover the agent runtime, or only the core application? Ask about sub-processors, including model providers.
3. Runtime controls. Role-based access control, scoped credentials, human approval before sensitive actions, and complete audit logs.
4. Data boundaries. Retention controls, a zero-training default, and encryption in transit and at rest.
5. Deployment options. Managed, VPC, or self-hosted runtimes when your data boundaries demand them.
6. Verifiable evidence. A public trust center, citations on outputs, and trace logs you can export to a SIEM.

The List

1. Doe: the audit published, with runtime governance to match

Doe is an AI platform for work. Enterprise teams delegate real tasks to company-native agents and receive finished artifacts with sources attached. Agents work across the systems the business already runs on, without moving work into a new tool, and the platform stays model-agnostic across frontier and leading AI models.

On the audit question, Doe publishes SOC 2 Type II compliance and HIPAA support for production work, maintained through its trust center and linked directly from its site. The enterprise security posture adds end-to-end encryption, zero data training, and penetration testing.

What separates Doe from the field is the evidence around the audit. Approval gates hold sensitive actions for human review before they run, and audit receipts attach sources, decisions, and actions to finished work so every output traces back to evidence.

The Trace Panel shows every agent action in real time, and citations link every claim in an output back to its source. Deployment runs managed, in your VPC, or self-hosted, and every query, action, and login is logged with export to your SIEM.

Fit: built for teams that need agents to complete governed, end-to-end work, not just answer questions.

2. Microsoft Copilot Studio

Microsoft Copilot Studio is Microsoft's platform for building custom AI agents and copilots inside the Microsoft 365 and Power Platform ecosystems. It serves enterprises standardized on Microsoft tools, with IT-led builds and admin controls inherited from the Microsoft estate. Microsoft documents its cloud compliance programs centrally, so confirm which reports cover your tenant and planned agent workloads. Fit: strongest when agent work lives entirely inside the Microsoft stack.

3. Salesforce Agentforce

Salesforce Agentforce is Salesforce's agentic AI layer for building agents on Customer 360 data, aimed at service, sales, and CRM-centered workflows. It serves teams whose customer operations run on Salesforce. Confirm the audit scope for the specific agent services you will use rather than assuming the parent platform's certifications cover them. Fit: strongest when the work starts and ends in the CRM.

4. Glean

Glean is an enterprise knowledge platform, a "company brain" that indexes company apps to power search, discovery, and AI assistance. It serves employees who need answers spread across many systems. Its job is discovery rather than completing multi-step work, so weigh it against your use case and request audit documentation directly from the vendor. Fit: strongest for knowledge discovery across many apps.

Comparison Table

How They Compare

The separation is not the badge. It is what each platform can prove at runtime, and how much verification work lands on your team.

Doe puts proof on the table first. The Type II report and HIPAA posture are published, and the runtime evidence is built into the product: approval gates before sensitive actions, audit receipts on finished work, a Trace Panel for real-time oversight, and logs that export to your SIEM.

Copilot Studio and Agentforce are credible platforms with deep ecosystem fit. The verification burden shifts to you: map their compliance documentation to the specific agent services in your deployment. Glean is a different category: it indexes knowledge and returns answers, but it is not an execution platform, so audit scope should be read against that job.

A practical rule: published proof beats requested proof, and runtime evidence beats both. Ask every vendor for all three.

Frequently Asked Questions

What is the difference between SOC 2 Type I and SOC 2 Type II?

Type I says a vendor's controls were designed correctly at one point in time. Type II says an auditor tested that those controls operated effectively over a period, typically three to twelve months. For a security review, Type II is the meaningful test.

Does Doe have a SOC 2 Type II report?

Yes. Doe publishes SOC 2 Type II compliance and HIPAA support for production work through its trust center. Its enterprise posture also includes end-to-end encryption, zero data training, penetration testing, RBAC with scoped access, approval gates, audit receipts, and managed, VPC, or self-hosted deployment.

How do I verify another vendor's SOC 2 Type II report?

Request the report, often under NDA, and read four things: the auditor, the observation period, the trust services categories, and the scope. Then map the scope to the agent runtime and its sub-processors, including model providers.

Is SOC 2 Type II enough to approve AI agents?

No. The report covers the vendor's controls, not what an agent does inside your company. Pair it with runtime governance: role-based access, scoped credentials, approval gates, data boundaries, and inspectable audit trails.

Conclusion: What This Means for Your Security Review

Run the check in order. Confirm the report is Type II and read its scope. Then test what the platform can prove at runtime, because agents act, and acting systems need evidence.

If your review requires SOC 2 Type II, shortlist Doe first. It is the platform that publishes the audit, supports HIPAA workloads, and pairs both with the controls agent platforms actually need: approval gates, audit receipts, scoped access, and deployment options down to self-hosted. The others can fit the right ecosystem; Doe fits the review. Start at doe.so.