AI agents and how to govern them: Tailscale, Saviynt offer solutions

The sales pitch for agentic workflows always has a risk caveat. Scaling agentic activity means making necessary upgrades in identity governance and authentication. With all those AI agents in the picture, someone has to set some rules.

As a release from Tailscale puts it, “as AI shifts from occasional chat usage to always-on coding assistants and autonomous agents, organizations need guardrails that keep up.” The Toronto-based startup joins those trying to address the need for advanced identity security that extends to agentic AI, with the launch of its Aperture product in open alpha.

Tailscale describes Aperture as “a governance layer that helps organizations adopt AI across both employees and agents.”

“Aperture provides centralized policy controls, audit-ready session histories, and safer handling of provider credentials, so teams can move fast with AI while reducing data leakage and improving compliance posture,” the firm says.

It aims to help organizations integrating AI agents move from experimentation to production, offering quick setup in an existing tailnet, centralized API key custody and identity-based usage tracking. The platform ties AI access and activity to a user or workload identity, so policies and audit trails “clearly reflect who initiated each request.” Comprehensive records capture LLM session histories and tool-call history.

“The pressure to adopt AI is forcing organizations to take risks they would never accept elsewhere,” says Avery Pennarun, CEO of Tailscale. “Security teams are being asked to approve AI deployments without clear attribution, consistent controls, or audit trails. Tailscale Aperture ties AI usage to identity and provides centralized logging and governance, so companies can adopt AI faster without creating an unmanageable security gap.”

Early customers in the alpha period include Oso, Cerbos, Apollo Research, and Cribl, all of which Tailscale Aperture is helping to “apply fine-grained authorization and operationalize AI governance using existing security and observability workflows.”

Per the release, Aperture supports both hosted and self-hosted AI endpoints at launch, including providers such as OpenAI, Anthropic, Google Gemini, Amazon Bedrock, OpenRouter and Vercel, alongside self-hosted endpoints. It also works with popular coding agents and agent frameworks where you can configure a custom base URL, including Claude Code, Codex, and Gemini CLI.

Saviynt joins Wiz Integration Network to fight NHIs

Saviynt has been enlisted to help cloud security firm Wiz combat the threat presented by non-human identities (NHIs), including rogue AI agents.

A release says the partnership provides “a unified view of risk by mapping Wiz’s cloud-native insights directly to Saviynt’s identity security workflows.” It will see Saviynt join the Wiz Integration Network (WIN), allowing customers to integrate Wiz into their existing workflows.

Key benefits include instant analysis and governance, risk-based prioritization to protect critical assets, proactive posture management that unifies visibility and immediate remediation, and complete registration and lifecycle management of NHIs and AI agents.

“Cloud and identity security are no longer separate disciplines. They are two sides of the same coin,” says Ehud Amiri, senior VP of product management at Saviynt. “By operationalizing Wiz’s cloud intelligence within Saviynt’s identity security framework, we are giving customers a way to secure every human, workload and AI agent without adding operational complexity.”

Article Topics

AI agents  |  authentication  |  cybersecurity  |  digital identity  |  identity management  |  identity security  |  non-human identities  |  Saviynt  |  Tailscale