From the Palo Alto Networks edition of November 6, 2026
Checkable AI docs review of Cortex XSIAM/XDR/AgentiX: 46 control findings last checked Nov. 3, with many identity, gateway and governance controls not addressed
Checkable AI · Independent AI security research / documentation review · Global (product docs; FedRAMP/IRAP notes for US/AU)
- Publisher
- Checkable AI
- Original publication
- 2026-11-03
- Source retrieved
- 2026-11-07
- Event date
- 2026-11-03
What happened
Checkable AI’s stack-check for Cortex XSIAM, Cortex XDR (including Agentic Endpoint Security) and AgentiX, last checked 3 Nov 2026, reports 46 findings on controls in scope. Supported examples include human approval for sensitive actions, agent remote scopes bound to user RBAC, HITL plan visibility, some runtime/third-party tool assurance via AES, AI discovery, asset inventory, posture reporting, and FedRAMP High/Moderate (plus IRAP notes). Large gaps called “not addressed” include per-agent issued identity, credential brokerage, OBO delegation to downstream MCP, local agent scopes in admin docs, kill switch/containment for in-flight agents, model/tool/data gateways, and most governance lifecycle workflows. Interpretation: documentation review—not a live red team—useful to temper vendor agentic SOC claims.
Why it matters
SLED evaluators considering Cortex agentic SOC features should use this checklist in architecture review: require human approval for containment, clarify that FedRAMP authorization does not equal full agent-identity controls, and demand local proof for AES guardrails that Checkable says appear mainly in blogs. Complements the archived Prisma AIRS Checkable review with SOC-platform coverage.
Evidence and measured results
Opened Checkable AI page stating 46 findings, last checked 3 November 2026. Findings cite Cortex docs URLs and vendor blogs. Explicitly not a test of a live deployment. Supported/not-addressed tallies by control area provided in tables.
Limitations and uncertainty
Docs-only methodology. Some “supported” items escalated for human settle (inference). AES details sometimes from blogs not admin docs. Page may change after Nov. 3 check date.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-07; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Buyers hear agentic SOC promises but lack an independent control checklist.
- Stakeholders
- CISO
- SOC architect
- Risk/compliance
- FedRAMP program office
- Discovery questions
- Which Checkable control areas are must-have for go-live?
- Is AES licensed or only XSIAM base?
- Who owns compensating controls for gaps?
- Value hypothesis
- Interpretation: transparent gap handling increases trust and focuses the PoC on demonstrable controls.
- Potential engagement
- Control-gap workshop mapped to Checkable findings plus a bounded AgentiX PoC.
- Unsupported claims
- That docs-supported equals operationally proven
- Full coverage of agent identity controls
- IRAP dating/version completeness
Pre-sales engineering
Role takeaway
- Fit
- Agencies deploying Cortex agents who need an evidence-based control matrix.
- Architecture and integration
- Cortex AgentiX + optional AES + external PAM/gateway for gaps.
- Prerequisites
- Access to Cortex docs and tenant
- AES entitlement clarity
- Security architecture owner
- Constraints
- Docs lag product
- Some findings blog-sourced
- Not a penetration test
- Security
- Default sensitive actions to approval; deny broad MCP write tools in production until reviewed.
- Proof of value
- Demonstrate sensitive-action approval
- Show agent action audit record
- Document compensating control for no agent workload identity
Delivery
Role takeaway
- Work
- Produce a gap workbook, configure approvals, and run a supervised agent investigation pilot.
- Dependencies
- SOC process owners
- IAM for compensating NHI controls
- Ownership
- SOC engineering owns Cortex config; security architecture owns gap decisions.
- Skills and adoption
- Analysts trained to read agent plans and approve/deny actions.
- Governance checkpoints
- Gap workbook signed
- Production MCP tool set approved
- FedRAMP boundary impact reviewed
- Acceptance criteria
- Proposed: human approval proven on one containment action
- Proposed: written compensating controls for top 5 gaps
- Proposed: AES license decision recorded
- Risks
- PoC scope creep across 46 findings
- Docs changing after check date
- Over-trusting FedRAMP as agent-governance proof
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Do not assume Cortex agents have first-class workload identities or gateway mediation; design external PAM/secret brokerage and approval gates accordingly. Treat AES admin-doc gaps as implementation risk.
Governance
Who approves, reviews and stays accountable for outcomes?
Map Checkable gaps to agency AI governance policies (ownership, registration, exceptions). FedRAMP High helps procurement but does not close agent lifecycle controls Checkable marks missing.
Security and privacy
What data, permissions and controls need testing?
Missing kill-switch and credential-brokerage documentation increases risk if SOC agents are given broad automation; keep sensitive actions approval-bound.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Clearer control expectations reduce wasted analyst time chasing agent features that are not documented.
Procurement
What should contracts, pricing and exit terms secure?
Attach control checklists to RFPs; treat FedRAMP as necessary but not sufficient for agentic SOC acceptance criteria.
Operating model
Which teams own the service once it runs?
Encourages a shared responsibility model: Cortex for SOC agents and approvals; agency IAM/governance for identity lifecycle gaps.
Publication history
- 2026-11-06Palo Alto Networks · Issue 087 resources
Stable resource ID: checkableai-cortex-xsiam-xdr-agentix-controls-46-findings-20261103