Tailscale Inc., a startup that helps enterprises establish secure network connections, today announced Aperture in open alpha mode to offer centralized policy control and auditability for artificial intelligence agents to reduce data leakage.

In the fast-moving world of AI, where usage is increasing by leaps and bounds, companies are finding themselves in a tight spot: Enterprise users and information technology teams rarely know exactly who, or what, they’re working with. This means their data is being offered to third parties and handled through networks they often don’t own and by parties they need to trust.

This realignment has radically changed how compliance works for security, and workplace AI usage standards mean that AI tools now must comply with a new set of standards for sensitive data. According to a recent study by Cyberhaven, 34.8% of corporate data that employees feed to AI tools and models was sensitive. A separate study led by the University of Melbourne with KPMG reported that 48% of all workers were uploading sensitive company data into public AI tools.

This behavior reveals a widespread problem among employees that is culture-wide and needs to be curtailed with safety-first training. At the same time, enterprise companies need to examine how they address the way that they think about and attend to where data resides and moves through their systems.

“The pressure to adopt AI is forcing organizations to take risks they would never accept elsewhere,” said Tailscale co-founder and Chief Executive Avery Pennarun. “Security teams are being asked to approve AI deployments without clear attribution, consistent controls or audit trails.”

Aperture bakes in ways to help teams move from experimentation to production AI workflows with identity-linked policy controls. Pennarun explained that it ties AI usage to identity to provide centralized logging and audit controls so companies can adopt AI faster with the watchword “safety,” creating manageable security.

Tailscale is working with partners such as Oso, Cerbos, Apollo Research PBC and Cribl Inc. to help customers apply granular authorization and functional AI governance using the company’s existing security tools.

Aperture supports both hosted and self-hosted AI endpoints at launch, and includes the most popular providers such as OpenAI Group PBC, Anthropic PBC, Google LLC’s Gemini, OpenRouter Inc. and Vercel Inc., alongside self-hosted endpoints. It also works with common coding agents and agent frameworks, including Claude Code, Codex and Gemini CLI.

“Aperture gives our developers easy access to company-approved AI models and gives us clear visibility into how those models are actually being used,” said Cribl CEO Clint Sharp.

Tailscale said the tool will be available at no additional cost during the alpha period across all plans, making it easier for all teams to evaluate identity-linked AI governance early. Pricing details will be announced as the Aperture approaches broader availability.

Image: SiliconANGLE/Microsoft Designer