AWS MSP partners auditing or renewing on or after January 1, 2027 will be scored against Validation Checklist (VCL) 8.0, AWS’s “AI-First” revision of the MSP Program audit. Audits through December 31, 2026 can still use VCL 7.1, which gives partners a short window to close gaps before the switch.
Key Takeaways
- AWS released VCL 8.0 on August 21, 2026. It becomes the only accepted checklist on January 1, 2027.[2]
- VCL 8.0 has 61 controls across six sections: Business, People, Governance, Platform, Security, and Operations.[2]
- 24 controls are net-new, 27 were updated (many to cover AI workloads), and 10 carried over unchanged. AWS also retired 15 VCL 7.1 controls by folding them into broader ones.[2]
- New mandatory controls include a Responsible AI policy, data governance for AI, prompt injection prevention, and an AI transformation roadmap.
- Evidence collected for VCL 7.1 may no longer be enough. Partners should map their evidence control by control against the new checklist, not against last year’s package.
This guide covers what changed, which controls are mandatory, what evidence auditors expect, and how to prepare. All requirements should be confirmed against the official AWS MSP Program Validation Checklist[1] and your AWS Partner Development Manager.
What Is AWS MSP VCL 8.0 and When Does It Take Effect?
VCL 8.0 is the eighth major version of the checklist AWS auditors use to validate partners in the AWS Managed Service Provider (MSP) Program. AWS calls it the AI-First release because it shifts the bar from proving cloud infrastructure management to proving AI-enabled delivery with measurable outcomes.[2]
| Date | What happens |
|---|---|
| August 21, 2026 | VCL 8.0 released and available for use |
| Through December 31, 2026 | VCL 7.1 remains valid for audits and renewals |
| January 1, 2027 | VCL 8.0 becomes the only accepted checklist |
The version that applies depends on when your next audit or renewal falls. MSP partners follow a 36-month cycle from their original award date: a Full Audit in Year 0, Performance-Based Renewals in Years 1 and 2, and another Full Audit in Year 3. Anything due from January 2027 onward is assessed under VCL 8.0.[2]
Passing the checklist also matters financially. Validated MSP partners can earn an AWS incentive on managed services revenue of up to 5% of gross billed revenue, capped at $0.5M, and only while they remain validated against the current checklist.[2]
What Changed Between VCL 7.1 and VCL 8.0?
VCL 8.0 is organized into six sections with 61 controls in total. According to AWS’s published Summary of Changes, the breakdown is:[1][2]
| Section | Total controls | New | Updated | Retained |
|---|---|---|---|---|
| Business | 9 | 5 | 2 | 2 |
| People | 8 | 3 | 5 | 0 |
| Governance | 10 | 5 | 5 | 0 |
| Platform | 8 | 3 | 3 | 2 |
| Security | 12 | 5 | 3 | 4 |
| Operations | 14 | 3 | 9 | 2 |
| Total | 61 | 24 | 27 | 10 |
AWS also retired 15 VCL 7.1 controls by consolidating them. For example, separate controls for role-based access, MFA, and IAM now sit under a single IAM control.[2] That consolidation is why renewing partners cannot rely on a line-by-line mapping from their previous evidence package.
AWS has publicly named only some of the individual controls behind these totals. For the rest, partners need to review the full checklist control by control rather than relying on the summary alone.
Which New VCL 8.0 Controls Are Mandatory vs. Recommended?
Four of the publicly named new controls are mandatory, and all four require written governance that most MSPs have never had to produce for an audit. The remaining named controls are recommended.[2]
| Control ID | Control | Level | What partners need to show |
|---|---|---|---|
| GOVP-002 | Responsible AI Policy and Framework | Mandatory | A formal framework covering fairness, accuracy, transparency, accountability, privacy, and safety, with at least one applied example |
| GOV-003 | Data Governance for AI | Mandatory | Policies for the quality, lineage, consent, retention, and sovereignty of data used by AI systems, including synthetic data |
| SECP-004 | Prompt Injection Prevention | Mandatory | Input validation, content filtering, and guardrails on every AI-facing endpoint |
| BUS-002 | AI Transformation Roadmap | Mandatory | A documented plan for moving from traditional to AI-first delivery, with milestones and dated progress |
| PLAT-004 | Agentic AI Platform | Recommended | Ability to build, deploy, and manage AI agents at scale |
| GOV-002 | AI Agent and Model Governance and Lifecycle | Recommended | A model and agent registry with versioning, approval gates, bias monitoring, and retirement steps |
| SEC-007 | Agentic Zero Trust and Blast Radius Containment | Recommended | Identity verification and data isolation between agents, plus a demonstration of containing a compromise |
| OPS-004 | GenAI and Agentic AI Observability | Recommended | Tracing of agent sessions, drift detection, decision audit trails, and cost monitoring |
| OPSP-003 | Toil Measurement and Reduction | Recommended | Toil tracked as a share of operations time, with an automation backlog and quarterly reduction |
| BUS-006 | Industry Vertical Specialization | Recommended | Depth in one industry vertical, or an existing AWS Industry Competency |
| PEO-003 | Dedicated AI Practice Lead | Recommended | A named person accountable for AI strategy and delivery quality |
| PEO-004 | Forward Deployed Engineers Team | Recommended | A defined engineering role embedded in customer engagements to drive AI use cases |
These controls apply even to MSPs with little AI work today. A roadmap and a Responsible AI policy assess strategy and readiness, not just current deployments.
Which Existing Controls Were Updated or Kept?
Core security and operations controls are still mandatory, and several now carry an AI requirement on top of what partners documented for VCL 7.1.[2]
| Control ID | Control | Level | Status in VCL 8.0 |
|---|---|---|---|
| GOVP-001 | Change Management Process | Mandatory | Updated: now covers AI workload changes alongside infrastructure and application changes |
| SEC-003 | Vulnerability Management | Mandatory | Updated: now covers AI model and agent dependencies |
| OPS-009 | Cloud Financial Management (FinOps) | Mandatory | Updated: now includes AI workload costs, such as GPU right-sizing and inference cost tracking |
| SEC-001 | Identity and Access Management | Mandatory | Retained |
| SECP-001 | Encryption and Key Management | Mandatory | Retained |
| PLAT-005 | Well-Architected | Mandatory | Retained |
| OPS-010 | Patch and Release Management | Mandatory | Retained |
For updated controls, evidence from a prior audit is a starting point, not proof of compliance. For retained controls, a check that evidence is current is usually enough.
Some controls can be waived if a partner already holds a relevant AWS Competency, such as MSSP, AI, DevOps, Migration and Modernization, or an Industry Competency.[2] Confirm exemptions before building new evidence.
What Evidence Do Auditors Expect?
Auditors look for evidence that a control works in practice, not just a policy that says it should. A strong evidence package is dated, version-controlled, and tied to a named owner for each control.
Typical evidence by area:
- Identity and access: access policies, MFA enforcement records, and documented access reviews.
- Encryption: configuration exports and key management records.
- Change and vulnerability management: change approval records, scan reports, and remediation tickets with closure dates, now including AI models and agents.
- AI governance: the Responsible AI policy, data governance policies for AI, a model or agent inventory, and records of guardrails and prompt injection testing.
- FinOps: cost reporting that covers AI workloads as well as standard cloud spend.
- Strategy: the AI transformation roadmap with dated milestones and progress.
The most common gap is a mismatch between written policy and what actually runs in production. Auditors cross-check documents against live configurations, so evidence should come from the systems themselves wherever possible.
How Should MSP Partners Prepare for VCL 8.0?
Start three to six months before your audit date,[2] and prioritize the mandatory controls that have no evidence history. Gaps found during an audit are recorded as action items, and AWS requires those to be closed before it approves a partner into the program.[3]
- Confirm which checklist applies. If your audit or renewal falls before December 31, 2026, decide whether to use VCL 7.1 or prepare for VCL 8.0 early.
- Map all 61 controls to current evidence. Work from the official checklist, starting with mandatory and new controls.
- Check Competency exemptions. Rule out waived controls before collecting evidence for them.
- Start with the new AI governance work. The Responsible AI policy, data governance for AI, and prompt injection prevention usually take the longest to document.
- Re-check updated controls. Change management, vulnerability management, and FinOps evidence now needs to cover AI workloads.
- Assign one owner per control. Each owner keeps evidence dated and current between audits.
- Submit the self-assessment on time. AWS expects the Self-Assessment Spreadsheet at least 30 days before the scheduled audit.[2]
Frequently Asked Questions
When does VCL 8.0 become mandatory?
VCL 8.0 becomes the only accepted checklist on January 1, 2027. Until December 31, 2026, partners can still be audited or renewed under VCL 7.1.[2]
How many controls are in VCL 8.0?
VCL 8.0 has 61 controls across six sections: Business, People, Governance, Platform, Security, and Operations. Of these, 24 are new, 27 were updated, and 10 were retained.[2]
Which new VCL 8.0 controls are mandatory?
The publicly named mandatory new controls are the Responsible AI Policy and Framework, Data Governance for AI, Prompt Injection Prevention, and the AI Transformation Roadmap. Existing mandatory controls such as IAM, encryption, change management, and vulnerability management remain in place.[2]
What happens if a mandatory control is not met?
Gaps found in the audit become action items. AWS requires them to be closed and evidenced before it approves entry into the MSP Program.[3]
Does VCL 8.0 apply to MSPs that don’t deliver AI services?
Yes. Controls such as the AI Transformation Roadmap and the Responsible AI policy assess strategy and governance, so they apply regardless of current AI workloads.
Can existing AWS Competencies reduce the audit scope?
Yes. Some controls are waived for partners holding relevant AWS Competencies, such as MSSP, AI, DevOps, Migration and Modernization, or an Industry Competency. Check the official checklist for the exact list.[1][2]
Audit Readiness Is an Evidence Discipline
VCL 8.0 rewards partners who can show, with dated and traceable records, that their controls work in production. VCL 8.0 makes a strong case for treating evidence as part of daily operations rather than something assembled immediately before an audit.
That principle holds well beyond cloud audits. At FreightOptics, we apply the same thinking to operational data: verified records, automated audit trails, and decisions you can trace back to the source. See how FreightOptics approaches audit-ready data.
