Innovation & Tech

Onyx Security and Glean Add Runtime Controls for AI Agents

Onyx Security and Glean's beta integration combines agent discovery with policy checks on prompts, model responses and tool execution as enterprise AI agents gain broader access.

By Daniel Okafor

4 min read

Onyx Security and Glean Partner to Secure Enterprise AI Agents
Onyx Security and Glean Partner to Secure Enterprise AI Agentscliff1126 / pixabay

What's News

  • Glean says its connector ecosystem spans more than 275 applications and data sources.
  • The Onyx Security integration is listed as beta in Glean’s connector catalog.
  • Discovery uses the Glean API; runtime enforcement uses supported Glean hooks.
  • Beta scope covers agent inventory, posture management, runtime visibility, prompts, model responses and tool execution.

By Daniel Okafor

Glean says its connector ecosystem spans more than 275 applications and data sources; Glean’s integration with Onyx Security is listed in beta.

Onyx Security and Glean announced a partnership linking Glean’s enterprise AI platform with Onyx’s secure AI control plane. The companies said the integration combines agent discovery, risk assessment, policy enforcement and runtime protection in a shared governance framework.

It addresses a basic gap. An agent may retrieve information under a user’s permissions and still lack controls over what it does next. A workflow can move from gathering context to sending a message, updating a record or invoking another tool, the companies said.

What does the partnership connect?

Glean provides a platform for building, deploying and operating enterprise agents. Its architecture includes two layers:

  • Glean’s Enterprise Context connects business applications and indexes company information, according to Glean.
  • Glean’s Enterprise Graph maps relationships among people, content and activity, the company said.

Permission-aware retrieval determines which information a user can access while the agent works, Glean said. The available context can link a customer account, its support history and the people responsible for that relationship.

Zubin Irani, VP of Partnerships at Glean, said: “As AI agents take on more complex work across the enterprise, they need access to the right context, with the right controls in place.”

Glean Protect already covers user and agent identity, source-system permissions, sensitive content, runtime threats and action alignment, the company said. It also produces audit events that can feed security monitoring systems. Onyx adds a separate enterprise-wide governance and risk layer around Glean agents and other supported AI environments, according to the partnership description.

How does discovery differ from runtime enforcement?\n

Discovery and runtime enforcement use different routes: discovery reads inventory and posture through Glean’s API, while runtime checks proposed actions at supported Glean hooks.

According to Onyx’s technical description, security teams can use the discovery connection to review agents and their access. That connection does not provide a mechanism for executing agent actions.

At runtime, Glean’s hooks provide supported checkpoints. Onyx uses these points to evaluate conversation content, prompts, responses and proposed tool executions against enterprise policies, the companies said. The controls seek to block unsafe requests and steer unintended behavior back toward the authorized task.

For example, access to internal account information may be appropriate for preparing a customer meeting, while sending confidential pricing outside the company may violate policy. An action checkpoint assesses the proposed disclosure before the tool executes it.

The partnership also addresses instructions embedded in external content. A document or tool response can attempt to redirect an otherwise legitimate workflow, the companies said.

Why must governance follow agent changes?\n

Onyx says security teams can discover Glean agents, understand their access and capabilities, establish ownership and business purpose, and assign risk classifications. Those records can feed enterprise AI governance and compliance policies, according to the company.

Continuous assessment covers changes to permissions, tools and data access. An agent approved for one workflow may later acquire a connector or capability that changes its risk profile, even if its name and owner remain the same, Onyx said.

Centralized audit evidence records agent activity and policy application. Security teams can use those records to investigate activity and review how policies were applied, the company said.

Gil Elbaz, Co-founder and Chief AI Officer at Onyx Security, said: “As agents gain greater access and autonomy, security teams need visibility and control over what they can do.”

He added: “By combining Glean’s enterprise context with Onyx’s security control plane, customers can govern agent activity consistently and confidently across their AI environment.”

What sits inside Onyx’s broader control plane?\n

Onyx positions its secure AI control plane across SaaS applications, cloud services, endpoints and custom code. The company says its platform combines agent observability, security posture management, governance and runtime safeguards.

Onyx lists several supporting capabilities:

  • Session replay and audit trails support investigations, according to the company.
  • Natural-language policy controls sit alongside conventional security controls.
  • Integrations connect security information and event management and security orchestration systems.

Onyx also offers gateway capabilities for AI and Model Context Protocol traffic. Its supervisory Guardian Agent is designed to identify risks and remediate issues across the platform, the company said.

Together, these capabilities bring Glean activity into a control framework that security teams can use across multiple AI systems, according to Onyx.

What does the beta label mean for deployment?\n

The integration appears in Glean’s connector catalog with a beta label, according to Glean. Its listed scope includes:

  • Agent inventory.
  • Security posture management.
  • Runtime visibility.
  • Controls on prompts, model responses and tool execution.

Runtime enforcement depends on supported checkpoints. Teams must consider checkpoint coverage for each workflow during deployment planning, the companies said.

For enterprises expanding agent use across departments, the partnership’s value depends on keeping the agent inventory current. It also requires carrying ownership and risk decisions into runtime policies and applying those policies where agents can affect business systems, the companies said.

The forward test is whether those controls remain aligned as each new connector expands an agent’s reach.

Source: unite.ai

Share this article:

More from Daniel Okafor

Daniel Okafor

Show full bio

Correspondent covering money, funding and market shifts around entrepreneurship, startups and venture funding at Business Bearings.

2,189 articles

Related articles

« Previous articleNext article »