OPAQUE Names Creator of Microsoft's Agent Governance Toolkit as Chief Platform Officer

Siddique joins OPAQUE to bridge the critical divide between agent governance policy and verifiable hardware enforcement, creating the first unified platform to make AI agents safe for high-stakes data and systems where real enterprise value resides

SAN FRANCISCO, June 3, 2026 /PRNewswire/ -- OPAQUE, the Confidential AI company, today announced that Imran Siddique, the creator of Microsoft's Agent Governance Toolkit (AGT), will join the company as Chief Platform Officer. He will lead OPAQUE's engineering organization and platform strategy, working with co-founder and CTO Rishabh Poddar, who continues to lead OPAQUE's research and cryptographic platform direction.

Enterprises are giving AI agents real reach — to read email, query CRMs and databases, execute transactions, and take actions with real-world consequences. The first problem is that agents are non-deterministic: they don't reliably do only what they're told. They call tools they shouldn't, reach data outside their task, and take actions no one approved. The second problem is that the same reach can be turned against you: a single piece of untrusted content — an email, a document, a support ticket — can contain hidden instructions that hijack an agent into quietly exfiltrating internal data. Unreliable governance is the single most-cited reason enterprises won't put agents near sensitive data or critical systems.

Most of the industry has tried to solve these challenges at the content layer — wrapping agents in guardrails that screen what goes in and what comes out. Guardrails help, but they're expensive and still leak: no filter reliably predicts a non-deterministic system, and none catches every hidden instruction. A durable fix adds two layers beneath the content layer. One decides what an agent may do — the governance policy. The other enforces it where software alone can't, and proves it holds: hardware enforcement that constrains what an agent can reach, stops data from leaving through unapproved paths even when an agent misbehaves or is manipulated, and produces independently verifiable evidence that the policies were enforced. AGT, which Siddique architected and open-sourced during his tenure at Microsoft, is the industry's leading framework and remains the sole toolkit that provides documented mitigation for all 10 OWASP Agentic AI Security Initiative risks. OPAQUE's Confidential AI platform provides the second. Siddique joins to bring the two together. For the first time, the creator of the industry's leading agent governance framework and the team that can prove enforcement held in hardware are the same company.

"The prize here is enormous — AI agents working directly on a company's most sensitive data and systems, where the real competitive advantage lives. But most teams are connecting agents to that on 'probably' — guardrails that probably catch the bad behavior, that probably stop the leak. You don't run a regulated business on 'probably.' It takes three layers: guardrails on the content, policy on what an agent is allowed to do, and hardware that enforces both and proves it held. Imran built the policy layer that the industry relies on. Together, we're turning 'probably' into proof," said Aaron Fulkerson, CEO of OPAQUE.

Siddique spent 18 years at Microsoft, leading engineering across Azure Core, Azure for Industries in sovereign deployments, Azure DevOps, and SQL Azure before creating AGT — the MIT-licensed framework for governing AI agents at runtime, available at github.com/microsoft/agt. He also founded and open-sourced Agent OS, Agent Mesh, and Agent SRE, spanning runtime governance, zero-trust networking, audit, and resilient multi-agent orchestration — projects he originated independently before contributing to Microsoft's open-source portfolio.

"Screening what goes in and out of an agent was never going to be enough. You need a layer that decides what it's allowed to do, and a layer that enforces and proves it in hardware. I built the first at Microsoft — AGT, the only framework that addresses all 10 OWASP Agentic AI risks by design. OPAQUE is the second," said Siddique.

The need is sharpest where the data is most sensitive — financial services, healthcare, and other regulated and sovereign environments, where an agent's mistake is a reportable event and "trust us" is not an acceptable control. The AGT on OPAQUE integration, with its first Evidence Pack, will be available alongside the Confidential Computing Summit on June 23, 2026, as the first deliverable of Siddique's tenure.

About OPAQUE

OPAQUE is the Confidential AI company. Born from UC Berkeley's RISELab, OPAQUE lets organizations run AI models, agents, and workflows on their most sensitive data with hardware-attested isolation and cryptographic proof that approved governance policies were actually enforced — evidence customers and regulators can verify independently. Customers include ServiceNow, Anthropic, Encore Capital, and leaders across financial services, insurance, and healthcare. Learn more at opaque.co.

Media contact:
Inkhouse for OPAQUE
[email protected]

SOURCE OPAQUE