Check out our new GigaOm Radar! Learn more

The Non‑Human Identity Crisis Your SOC Can't See

The alarms start at 3:17 a.m. By sunrise, a Fortune 500 has lost its CRM data, billing records, and a chunk of proprietary source code. Not to ransomware. Not to a nation-state zero-day. To a single stolen token tied to a "harmless" SaaS service account. A background bot. A non‑human identity no executive has ever heard of.

This wasn't a misconfiguration story. This was the day the industry discovered that "shared responsibility" quietly excluded the most dangerous identities in the stack. Now legal is blaming the vendor, the vendor is blaming the tenant, and regulators want blood. The blast radius is global. The playbook is useless. And the old cloud security model just died on a Tuesday.

The Bots Are Running the Show

Cloud and SaaS don't really run on users; they run on robots. These "non-human identities" are the logins for services and scripts: service accounts, automation tokens, API keys, OAuth apps. They move data, deploy code, sync systems. Customers treat them as plumbing and hide them inside CI/CD pipelines and dusty scripts. Providers tuck them behind "one-click" integrations and glossy marketplace tiles.

On paper, almost every security story is about user MFA and configuration checklists. In reality, the keys to the kingdom are long‑lived, over‑privileged tokens that almost nobody monitors. The industry made a deliberate choice: ship integrations fast, juice ecosystem growth, and hope nothing bad happens. Governance for these identities never caught up and now the bill is coming due.

The Kill Chain Nobody Sees

The kill chain starts somewhere boring: a single service account or OAuth app tied to a popular SaaS integration gets popped. Maybe the token leaks from a CI/CD log, a support export, or a compromised partner inbox. That one "bot account" often has broad, inherited permissions across CRM, storage, and internal tools.

From there, the attacker rides built‑in trust. They use that token to fan out into connected tenants, dev and prod environments, maybe even partner instances, all without tripping classic user-based alerts. They clone tokens, abuse refresh tokens, and drop background jobs so they can come back whenever they want.

Logs are split between the customer's SIEM and the provider's black box. Each side sees half a ghost. When it blows up, both legal teams point to the shared responsibility diagram, and every dashboard suddenly looks designed for plausible deniability, not real accountability.

The "Misconfiguration" Scapegoat

Post‑mortems love one phrase: "tenant misconfiguration." The script writes itself; the customer granted excessive scopes, skipped a best practice, clicked through one scary warning. Case closed.

It sounds neat, but blaming tenants for non‑human identity abuse is like blaming drivers for brake failures. Technically true, strategically useless. Cloud and SaaS providers design the default scopes and permission models. They decide which logs exist, what shows up in admin consoles, and whether strange bot behavior is even visible to customers.

When a service account can quietly read everything and no one can see it doing so, that's not just "user error." As long as providers can hide behind "you clicked accept," they can ship risky defaults, weak telemetry, and vague diagrams—and leave customers holding the bag when the robots go rogue.

Redefining the Model: From Fragile Integration to Scalable Capability

A real "shared identity responsibility" model starts by admitting the current one is broken. For the enterprise CTO, this isn't just about avoiding a headline-grabbing breach; it's about solving the fragility that prevents a team from maturing. When a single "smart guy" has to manually monitor every service account and "dusty script" just to prevent a total system collapse, you don't have an automation strategy—you have a scaling bottleneck.

To move beyond this, we need to shift from out-of-band observation (searching for "ghosts" in logs after the fact) to in-band architectural enforcement.

A Three-Stage Framework for Identity Maturity

Instead of abstract claims of "better security," enterprises should demand a framework that enables new work through reliability:

Stage 1: Exposing the Managed Inventory  

Providers must provide a first-class, tenant-visible inventory of every non-human identity, including those the provider quietly manages behind the scenes. You cannot govern what you cannot see, and "hidden" plumbing is where fragility lives.

Stage 2: Validation of Intent  

We must separate technical blast radius from business intent. The provider must own the default controls—ensuring "god mode" scopes are technically unconfigurable—while the customer defines the intent: which bot is allowed to touch which data, and for how long.

Stage 3: Secure-by-Default APIs  

Unsafe identity patterns should fail at the API level. If a token allows "impossible travel" or sudden cross-tenant fan-out, the system should block the action automatically rather than just firing a silent alert into a black box.

Interrogating the "One-Click" Trap

Vendors have a clear incentive to prioritize "one-click" integrations: it boosts ecosystem growth and lowers the barrier to entry. However, this "efficiency" is a byproduct that creates long-term complexity for the customer. Real capability is found when you automate the necessary security steps so your team doesn't have to, enabling them to focus on high-value AI and data projects instead of auditing "harmless" SaaS accounts.

If a tenant cannot see or control a non‑human identity, the provider must own the risk by default. It is time for buyers to get blunt in their RFPs.

Six Hard Questions Every Buyer Should Ask


Here are blunt, high-impact questions for RFPs and vendor meetings. While they may be a little provocative, I have worded them in this way to ensure the intent of the message is clear. These are designed to move beyond marketing "efficiency" and focus on operational capability and risk ownership:

1. Inventory Transparency  

"Provide a live, tenant-visible inventory of every service account, workload identity, and OAuth token your platform creates in my environment—including those you manage internally. If I can't see it, will you contractually assume 100% liability for its misuse?"

2. Default Fragility  

"Why are your default integration scopes set to 'broad' or 'administrative' by default rather than 'least privilege'? Is this a technical limitation of your architecture or a choice to prioritize 'one-click' onboarding over my organization's stability?"

3. In-Band Enforcement  

"Does your platform support 'secure-by-default' controls that cause unsafe non-human identity patterns (like scope escalation or impossible travel) to fail at the API level, or do you only provide 'out-of-band' alerts that my team has to investigate?"

4. Incentive Alignment  

"How does your pricing model for telemetry and logging affect my ability to monitor non-human identities? If I have to pay extra for the logs required to see what your robots are doing, aren't you incentivized to keep me in the dark?"

5. Team Scaling  

"Show me how your identity management framework reduces the manual overhead for my security engineers. Does this system allow 'one smart person' to manage a hundred automated projects, or does it add another layer of 'fragile' complexity they have to babysit?"

6. The "Shared" Reality  

"In the event of a breach involving a compromised service account that was created by your platform with a default 'god-mode' scope, what specific percentage of the liability do you accept? Or does your 'Shared Responsibility' model end the moment I click 'Accept'?"

The Winning Move

Some vendors are about to discover they picked the wrong side of history. If you're still selling pretty dashboards while hiding non‑human identity complexity, you deserve to be on the losing end of the next mega‑breach.

The smart move is the opposite: treat "shared identity responsibility" as your wedge. Expose every awkward detail about your own service accounts and tokens. Give customers raw, sometimes uncomfortable visibility, and you will earn board‑level trust.

The next great cloud breach will be blamed on a customer click. History will blame the industry for pretending the bots were out of scope.

Discover what GigaOm can do for you

Independent, practitioner-led research that helps you choose, prove, and sell enterprise technology with confidence.

[Local dev] blog_closing_band_hubspot_form_id is not configured — the embedded HubSpot form would appear here.