OpenText and ThreatLocker Set Guardrails for Enterprise AI

What changed

OpenText is positioning itself as the trusted data and context layer for enterprise AI, CEO Ayman Antoun said at OpenText World 2026 in Orlando on October 20. The company says only about 20 percent of enterprise information is currently usable in agentic AI workflows, and is building its strategy around governed data across information management, business networks, IT operations and cybersecurity.

ThreatLocker also launched Zero Trust AI Control, a system designed to restrict what AI agents can access, execute and connect to. Its controls can limit application execution, file access and network connections, including when an agent invokes another application, script or process.

Why it matters

The two announcements point to the same practical problem: AI agents may be capable of taking action, but capability is not the same as permission.

OpenText’s bet is that enterprise AI will depend less on owning every model and more on making company information accurate, accessible and governed. Its partnership with Cohere makes that division clear: OpenText provides the data and context layer, while Cohere supplies models and orchestration through North.

That could move AI work toward the unglamorous preparation that determines whether a pilot survives contact with a real business: cleaning data, connecting systems and deciding who or what can use it. It also gives solution providers possible work around integration, governance and production deployment. How that opportunity fits with OpenText’s existing managed-services partners remains unresolved.

ThreatLocker’s approach puts the control outside the agent itself. An agent could analyze one approved folder and produce a report while remaining blocked from unrelated files or unauthorized destinations. That is a more concrete boundary than simply asking an AI system to follow instructions.

The outlook is constructive but unproven. If these controls work as described, enterprise teams may gain a clearer path from experimentation to constrained, auditable deployment. Pricing, adoption and technical performance are still unknown.