Capability · AI & Agentic Security

Securing non-human identities and AI agents.

AI agents, service accounts and API keys all sign in with credentials. Demand Cyber places non-human identity and workload identity engineers who inventory them, give each one an owner and limit its access to what the task needs.

What is agentic security?

Agentic security controls what autonomous AI agents can access and do on your behalf. It covers the agent's identity, its permissions, the tools it can call and the guardrails on what it produces.

At a Glance

Where do guardrails sit in an agentic workflow?

At every step between the request and your data. An agent should only reach the tools and data its task needs, and high-risk actions should wait for a person to approve them.

01RequestA person or a trigger asks the agent to do something.
G1Input controls
02AgentThe model plans the task and picks tools.
G2Least-privilege scopes
03Agent identityA non-human identity with its own credentials.
G3Approval for high-risk actions
04Tools & APIsEmail, tickets, code, payments, search.
G4Logging and review
05Data & systemsCustomer records, files, production systems.
G1Input controlsFilter prompts and inputs for injection before the agent acts.
G2Least-privilege scopesThe agent identity gets only the scopes this task needs.
G3Approval for high-risk actionsPayments, deletions and external sends wait for a person.
G4Logging and reviewEvery action is logged against the agent identity and reviewed.

What are non-human identities?

They're the service accounts, API keys, tokens and workload identities that machines and agents use to sign in. They're often created quickly and reviewed rarely, which makes them an easy way in.

Every new agent adds more of them. Governing them uses the same disciplines as human identity: an owner, least privilege, regular review and a clean way to retire them.

What do your specialists deliver?

An inventory and owner for every non-human identity, least-privilege access for agents and workloads, and a clean way to rotate and retire credentials. The detail depends on how far along you are.

AreaWhat the specialist delivers
Non-human identity governanceInventory, ownership and lifecycle for service accounts, keys and tokens
Agent permissionsLeast-privilege scopes for agents and the tools they call
Workload identityService-to-service authentication with SPIFFE/SPIRE or Entra Workload ID
OAuth app governanceConsent review, token hygiene and removal of unused app grants
Agent guardrailsApproval steps for high-risk actions, logged against the agent identity
And MoreAsk us about other areas of non-human identity work.

Which non-human identity tools do you cover?

Specialist tools such as Astrix, Entro, Oasis and Aembit, plus workload identity in SPIFFE/SPIRE and Microsoft Entra Workload ID.

Several of these tools are being brought into SailPoint, Cisco and Cyera, so non-human identity is moving into the governance platforms many teams already run.

Which non-human identity roles can you fill?

Engineers who govern agent and workload identities, and the architects and authorisation engineers who design how they get access.

Non-Human Identity EngineerWorkload Identity EngineerIdentity Architect, Workloads & AgentsAuthorisation EngineerAI Security EngineerAnd More

How do you vet AI and agentic security specialists?

Every AI and agentic security specialist we put forward has passed four screening stages run by CISSP-certified practitioners, including a hands-on deep-dive in a live sandbox.

Of everyone we source, 42% pass the credential and background check, 18% pass the sandbox deep-dive, 6% pass the practitioner panel and about 1% are cleared for placement. See how we vet or view a sample candidate report.

Can I hire an AI security specialist on contract or permanently?

Both. A contractor can build your first non-human identity inventory and guardrails, and a permanent hire can run them from there.

Permanent hires are fee-based, with a replacement guarantee. Contract Placement · Permanent Hire · Advisory & Workshops

Questions

Common questions.

We're early with AI. Where should we start?+

With an inventory of the agents and non-human identities you already have. A fixed-scope advisory review can produce one, with a prioritised plan.

Do your specialists build AI products?+

No. They secure the identities, permissions and credentials those products use.

Is this separate from our identity programme?+

It works best inside it. The same governance that covers staff accounts should cover agents and workloads.

What if I don't have the budget for an onshore resource?+

Our Global & Blended Teams option pairs an onshore lead with vetted engineers in our talent hubs in the Philippines, Vietnam and Singapore. You keep a local point of contact at a lower blended day rate, and every engineer passes the same practitioner vetting. See Global & Blended Teams

Need an AI security specialist?

Book a 30-minute scoping call. You'll speak to a practitioner, not a recruiter.