Teleport: Agentic AI Forces an Identity Rethink

The rise of agentic AI is redefining the concept of a “user,” and that opens some big security questions that lie beneath the promise of agents and automation.

Enterprises are being pressured to adopt AI rapidly, including agents—but they’re also being squeezed to tighten down security. It’s a tough combination of tasks, considering agentic AI amounts to a new herd of users, possibly created without human scrutiny and potentially carrying credentials to access many data resources.

This is, of course, a recipe for trouble. Identity-based specialist Teleport today is responding to this new challenge with the launch of Teleport Agentic Identity Framework, a platform aiming to let enterprises apply that unified identity layer right away. Teleport argues the answer requires building a unified identity foundation that oversees human and nonhuman privileges across all systems.

Identity as Battleground

Identity is becoming the next big battleground for security vendors. The most dramatic move in this direction came in July when Palo Alto Networks bid $25 billion for CyberArk, an identity security and privileged access management (PAM) firm competing with the likes of Okta and SailPoint. Given Palo Alto’s strategy of platformization, the implication is that a security “platform” is no longer complete without identity management.

New identity-security contenders are gaining ground, too. ConductorOne, founded in 2020, raised a $79 million Series B last year. The company says its goal is to “rebuild identity security from the ground up for the modern workforce.”

Teleport, likewise founded in 2020, has raised $165 million in three rounds, the latest in 2022. The company’s platform provides a unified identity layer, consolidating the many credentials a user (human or not) might have across multiple data resources, whether on-premises or in the cloud. It’s also begun emphasizing the need to track a real-time “chain of custody”—that is, a user’s activity across multiple platforms.

Teleport boasts more than 700 customers, initially drawn from the ranks of SaaS services; banking and crypto trading outfits; and gaming companies. Increasingly, though, Teleport has been drawn into the AI infrastructure world, securing GPU datacenters and AI model building, and it was early in pointing out how the Model Context Protocol (MCP) could create accidental security holes.

"Opinionated" Is Good

The Teleport Agentic Identity Framework is production-ready, meaning it doesn’t require technology that’s being promised for the future. Teleport certainly intends to continue augmenting the framework with upcoming products. The point of the framework is to provide a roadmap enterprises can apply immediately.

That’s important because Teleport believes enterprises need to do a heavy lift, rethinking identity in a way that unifies all users (human or otherwise) across all systems and data sources. That foundation is necessary, because simply overlaying new AI models and agents atop existing security will only magnify the existing identity and access vulnerabilities, the company argues.

A key aspect of the framework is that it’s “opinionated,” said Diana Jovin, Teleport’s chief marketing officer. Enterprises are drowning in multiple waves of AI decisions, from choice of LLMs to the many scenarios for deploying agents. Teleport’s strategy is to provide direction rather than paralyzing customers with options.

That seems to be resonating. Teleport ran an unscientific experiment during KubeCon in Atlanta last fall. On Day 1, they focused their conversations on their products and capabilities. The next day, Teleport’s crew started espousing the upcoming identity framework. The increased enthusiasm was palpable.

“Boy, people leaned in. They are hungry for guidance,” Jovin said. “There is some appetite for this assembly of information and resources.”