What the IaaS questions on your application are really testing
If your renewal application now has a cloud page, it's asking something more specific than "do you use MFA?" Underwriters want to know three things. Could one stolen credential become a full-tenant compromise? Could one misconfigured bucket trigger a breach notification? Could an attacker delete your backups before encrypting production? Those are the loss scenarios behind cloud claims, and the newer questions map straight onto them.
Wording on cloud security requirements for cyber insurance varies by market. A London-market wholesale form, a US admitted carrier's supplemental, and a proposal form from a European, Gulf or Singapore insurer will all phrase things differently. They still probe the same six areas: privileged identity, public exposure, backup isolation, posture monitoring, logging and asset inventory. Weak answers in any of them can bring subjectivities, ransomware sublimits, higher retentions or a decline.
Below are eight gaps that commonly stall a submission, each with a fix and evidence an underwriter is likely to accept. A "yes" on the form is a representation. If it's wrong, you could face a coverage dispute after a loss. So answer from configuration exports, not memory, and have your broker review anything you're unsure about. For wording questions, bring in coverage counsel.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Gaps one and two: weak root protection and standing admin rights
Red flag: the top-level account uses SMS codes, has no MFA at all, or still has access keys. Root account MFA on AWS is usually the first thing a reviewer checks, because the root user can bypass most other controls. The same logic covers Azure Global Administrators and Google Cloud super admins in Cloud Identity or Workspace. Provider-enforced MFA closes part of the gap. It doesn't make the MFA phishing-resistant, and it does nothing about leftover programmatic keys.
Do this instead:
- AWS: Register at least two FIDO2 security keys on the management account's root user and store them in separate locations. Delete any root access keys. Use centralized root access management in AWS Organizations to remove root credentials from member accounts.
- Azure: Put two cloud-only break-glass accounts on FIDO2 passkeys. Require phishing-resistant authentication strength for admin roles through Conditional Access, and alert on every break-glass sign-in.
- Google Cloud: Enforce security keys for super admins through 2-Step Verification policy. Keep those accounts out of daily use and grant Organization Admin to separate named identities.
Red flag: a dozen people hold Owner, Global Administrator or AdministratorAccess permanently. Standing privilege turns any phished engineer into a tenant-wide incident. Move to just-in-time elevation: Entra Privileged Identity Management on Azure, Privileged Access Manager on Google Cloud, and on AWS either temporary elevated access workflows in IAM Identity Center or a third-party PAM tool. For most mid-size estates, aim for single digits of permanent admins, ideally just the break-glass accounts.
Evidence that satisfies: the AWS IAM credential report showing root MFA active and no active root keys, the Entra authentication methods registration report, the Google Admin console 2SV enrollment report, and a role-assignment export counting permanent versus eligible admins.
Gaps three and four: public storage and open management ports
An underwriter can often confirm this misconfiguration without asking you. Many carriers and MGAs run outside-in scans during underwriting, so an exposed RDP port or public bucket may appear in the quote conditions before you've finished the form.
Red flag: object storage is private only because nobody has made it public yet. That's a habit, not a control. Enforce the setting at the highest level:
- AWS: Turn on S3 Block Public Access at the account level and back it with a service control policy so member accounts can't switch it off.
- Azure: Disable blob anonymous access on every storage account, and enforce it with an Azure Policy assignment at the management group.
- Google Cloud: Apply the public access prevention organization policy constraint, along with domain-restricted sharing.
If you genuinely need public content, such as website assets, put it in a separate, clearly labeled account or project behind a CDN and declare it on the form.
Red flag: security groups or firewall rules allow SSH (22) or RDP (3389) from 0.0.0.0/0. Exposed remote access is a common ransomware entry point, and underwriters treat it that way. Replace it with identity-aware access: AWS Systems Manager Session Manager, Azure Bastion or Just-in-Time VM access in Defender for Cloud, or Google Cloud Identity-Aware Proxy TCP forwarding. Database admin consoles and Kubernetes APIs need the same treatment.
Evidence that satisfies: exports of the SCP and policy assignments, a CSPM report showing no public buckets and no internet-open management ports, and, if you have one, a recent external attack surface scan that matches what the carrier sees.
Gap five: backups that sit inside the same blast radius
"We use the provider's backup service" doesn't answer the ransomware question. If the admin credential that controls production can also delete the backup vault, the backups are inside the blast radius. Isolation means three things: immutability the attacker can't undo, separation into a different account or subscription, and a restore you've actually tested.
| Platform | Immutability control | Separation approach | What to show the underwriter |
|---|---|---|---|
| AWS | AWS Backup Vault Lock in compliance mode | Copies to a dedicated backup account in the organization, ideally in a second region; logically air-gapped vaults where available | Vault Lock configuration, cross-account copy job history, last restore test |
| Azure | Immutable vault (locked), soft delete always-on | Multi-user authorization via Resource Guard in a separate subscription or tenant | Vault immutability state, Resource Guard assignment, restore test record |
| Google Cloud | Backup vaults with enforced minimum retention; bucket retention locks for exported data | Backup vault in a separate project with distinct IAM | Vault retention settings, project IAM export, restore test record |
Two practical points. First, compliance-mode locks can't be shortened once set, so choose retention periods with your legal and privacy team. Data-protection rules such as the GDPR, India's DPDP Act and Brazil's LGPD limit how long you can keep personal data. Second, SaaS data (Microsoft 365, Google Workspace, Salesforce) usually needs a separate backup product, and many supplementals ask about it as its own question.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Gaps six and seven: no posture monitoring, and logs an attacker can switch off
Red flag: misconfigurations get found during annual pen tests or after incidents. Carriers increasingly ask whether you continuously monitor cloud configuration. For any company with more than a handful of accounts or subscriptions, cloud security posture management (CSPM) is effectively expected. Native tools are a reasonable start. A dedicated platform makes sense once you run more than one cloud or need agentless workload and identity analysis.
| Option | Best fit | Indicative 2026 cost (USD) |
|---|---|---|
| Native: AWS Security Hub, Microsoft Defender for Cloud CSPM, Google Security Command Center | Single-cloud estates, small-to-mid teams | Roughly a few hundred to several thousand per month, driven by resource count and checks |
| Third-party CNAPP: Wiz, Orca, Palo Alto Prisma Cloud, CrowdStrike Falcon Cloud Security | Multi-cloud, mid-size to large | Often tens of thousands to low six figures per year for mid-market; enterprise contracts higher |
| MSSP-managed CSPM | Lean security teams needing triage | Tool cost plus a managed fee, commonly a few thousand per month |
These are typical 2026 market ranges. They vary by contract, region and workload volume, so confirm with vendors. Underwriters don't need a perfect score. They want a running tool, tracked critical findings and a stated remediation time for high-severity issues.
Red flag: audit logs live in the same account an attacker would compromise. If the attacker's first move can be disabling CloudTrail or deleting a Log Analytics workspace, both your investigation and your claim suffer, and forensic costs rise when there's nothing to reconstruct. Send an AWS organization trail to a dedicated log archive account with S3 Object Lock. Export Azure activity and Entra sign-in logs to a workspace or storage account in a separate subscription with restricted access. On Google Cloud, route audit logs through an organization-level sink to a bucket with a locked retention policy. Twelve months of retention is a common baseline, though regulations in your sector may require more.
Gap eight: accounts, subscriptions and projects nobody declared
This is the gap that hurts most after a loss. Picture a developer's personal AWS account holding a copy of production data, an Azure subscription inherited from an acquisition, or a contractor's GCP project billed to a corporate card. None of these appear in your CSPM, and none follow your MFA or backup controls. If a breach starts there, an insurer may argue your representations about cloud controls were inaccurate.
Do this instead: build the inventory from billing data, not org charts. Reconcile corporate card and expense reports against cloud provider invoices. Bring stray AWS accounts into your organization, move Azure subscriptions under your management group hierarchy, and attach Google Cloud projects to your organization node. Anything you can't bring under control, shut down or disclose on the form with the controls it actually has.
Data location deserves its own column in the inventory. Many proposal forms ask where personal data is stored, and the answer matters more if you operate under localization or transfer rules: Europe's GDPR transfer regime, Saudi Arabia's PDPL, China's PIPL, or sector rules from regulators such as the Reserve Bank of India or the Monetary Authority of Singapore. Regions also affect breach-response costs and which regulators you'd notify.
Evidence that satisfies: a list of accounts, subscriptions and projects with owner, environment, region and data classification, plus a statement of how often you reconcile it. Quarterly is a defensible cadence for most mid-size companies.
Assembling the evidence pack before your submission
Start eight to twelve weeks before renewal. Most fixes are configuration changes, not purchases, but change control and testing take time, and the backup lock needs a retention decision that legal must sign off on.
| Gap | Typical effort | Evidence artifact |
|---|---|---|
| Root/global admin MFA | Days | Credential report, auth methods report |
| Standing admin privilege | 2 to 6 weeks | Role-assignment export, PIM/PAM logs |
| Public storage | Days to 2 weeks | SCP/policy export, CSPM report |
| Open management ports | 1 to 4 weeks | Firewall export, external scan |
| Backup isolation | 2 to 6 weeks | Vault lock settings, restore test |
| Posture monitoring | 1 to 8 weeks (procurement-dependent) | CSPM dashboard summary, remediation SLA |
| Tamper-resistant logging | 1 to 3 weeks | Org trail/sink config, retention lock |
| Undeclared accounts | 2 to 6 weeks | Reconciled inventory |
Effort estimates are rough planning figures; they depend on estate size and change-control rules. Package the artifacts as a two-to-four-page cloud controls summary with screenshots in an appendix, and give it to your broker with the application. A clear summary tends to cut follow-up questions and gives the broker something concrete to negotiate with.
If time is short, fix root MFA, public storage and backup isolation first. Those three are the most likely to produce a decline or a ransomware sublimit. For anything you can't finish before binding, don't stretch the truth on the form. Disclose it with a dated remediation plan, and ask your broker whether the carrier will accept it as a post-bind subjectivity.