Rig Security Bets on Identity Graphs as Agentic AI Turns Access Into the Real Attack Surface
With $12 million in funding and a graph-based identity pitch, Rig Security is entering a market where the hardest question is no longer what an AI can do, but whose authority it is using when it does it.
Rig Security has emerged from stealth with a claim aimed at one of the messiest problems in agentic AI: identity. In systems where software agents can call tools, move through workflows, and act with limited human supervision, the security issue is not just model output. It is whether a machine action is being carried out under a human, service, or agent identity, and whether defenders can still trace and constrain that path.
That is why the company’s “identity dependencies graph” matters as an idea, even before any product claims are independently proven. In practical terms, the concept points toward mapping relationships among users, permissions, sessions, and trust links so operators can inspect delegated access rather than assume that a login event tells the whole story.
Fast Facts
- Rig Security emerged from stealth with $12 million in funding.
- The company is positioning its work around identity risks tied to agentic AI.
- Its pitch centers on an “identity dependencies graph” for separating legitimate user activity from AI-driven activity.
- Agentic systems can create accountability gaps when delegated access is broad or poorly tracked.
- Open guidance for AI agents increasingly emphasizes explicit authorization, least privilege, and auditability.
TECHCROOK
The technical challenge here is closer to IAM than to model safety. An agent that can act on behalf of a person may inherit that person’s access, but the security value of the action depends on whether the delegation is narrow, time-bound, and visible. If the same credentials are reused across human and machine workflows, attribution gets muddy fast.
That is where graph-based identity analysis can be useful. A graph can show transitive trust, standing privileges, and session relationships that do not appear in a single sign-in record. From a defensive perspective, that helps answer questions such as: who authorized the agent, what can it reach, and what should have been off limits?
Still, the important word is “can,” not “will.” A graph-based system is an architectural control, not a guarantee. It may improve visibility, but its real-world value depends on how identities are issued, how permissions are scoped, how credentials are bound, and whether high-risk actions still require separate approval. The same is true for agentic security more broadly: the best controls are usually the boring ones, such as least privilege, short-lived credentials, strong audit trails, and continuous monitoring.
At the time of writing, the available information supports a risk analysis, not a proof that any one platform can reliably solve agent attribution in every environment. That distinction matters, because agentic AI failures often come from delegation errors, not from the model “misbehaving” in a vacuum.
Conclusion
Rig Security’s entrance reflects a larger shift in cyber defense: as AI systems move from suggesting actions to executing them, identity becomes the battlefield. The lasting lesson is simple. If an agent can borrow authority, then security teams must be able to see exactly where that authority came from, what it covers, and when it should stop.
TECHCROOK
hardware security key: A hardware security key is a simple, widely available way to strengthen login protection for email, SSO, admin portals, and other sensitive accounts. It adds a physical factor that is harder to phish than passwords alone. For teams focused on identity and delegated access, it is a practical baseline control alongside least privilege, short-lived credentials, and audit trails.
WIKICROOK
- Agentic AI: AI systems that can plan and take actions with limited human supervision.
- Identity Dependencies Graph: a relationship map of identities, permissions, sessions, and trust paths.
- Least Privilege: a principle that limits each identity to only the access it needs.
- Delegated Access: permission granted to one identity to act on behalf of another.
- Non-repudiation: a property that helps make actions attributable to a specific identity, reducing the ability to deny responsibility.
