Immutable Backup Requirements for Cyber Insurance: A 10-Step Readiness Playbook
Ransomware supplements ask whether your backups are "offline or immutable." This 10-step sequence turns that checkbox into evidence an underwriter will accept and a claims adjuster will find hard to pick apart.
What 'Offline or Immutable' Actually Means on an Insurer's Form
Most ransomware supplements ask some version of one question: are your backups offline, air-gapped or immutable? Ticking "yes" takes a second. That tick is also a representation on your application. Say a claim later shows that an attacker with domain admin rights deleted every restore point. The insurer may then have grounds to dispute coverage, reduce the payout or reprice you sharply at renewal.
Carriers rarely publish their exact immutable backup requirements, and the wording differs between insurers and markets. In practice, underwriters are checking four things:
- Deletion resistance: can someone holding stolen production credentials delete or encrypt your backups, or shorten how long they're kept?
- Protection window: how many days of clean restore points survive if the attacker was in your network for weeks before acting?
- Isolation: does at least one copy sit outside your main network and your main identity system?
- Proven recovery: have you restored from those copies recently, and can you show it?
Do the ten steps below in order. Each one ends with the evidence to keep. A backup control you can't document counts for little at underwriting and for even less once a claim is filed.
Map the Systems That Must Come Back First
Step 1: Inventory and tier what you protect
List every workload that holds business-critical data. That means virtual machines, physical servers, databases, file shares, SaaS data (Microsoft 365, Google Workspace, Salesforce) and the identity systems themselves. Identity is the item that most often gets left off the list. If you can't rebuild Active Directory or your Entra ID configuration, you can't restore much else.
Then sort each workload into a tier. Tier 1 is anything that stops revenue or operations within a day, which usually means ERP, identity, core databases and production file shares. Tier 2 starts to hurt within a week. Tier 3 can wait.
Step 2: Set recovery targets per tier
Write down a recovery point objective (RPO) and a recovery time objective (RTO) for each tier. Get the business owners to sign them off, not just IT. Take a hypothetical 250-person manufacturer that sets Tier 1 at a 4-hour RPO and a 24-hour RTO. That one decision sets its backup frequency and roughly how much immutable storage it needs. It also sets what its restore tests have to prove.
Evidence to keep: the tiered inventory and the signed RPO/RTO table, both dated.
Cut the Backup Platform Off From Production Identity
Step 3: Separate backup credentials
A common way backups fail in a ransomware incident has nothing to do with software bugs. The attacker takes over a domain admin account and uses it to log in to the backup console. Close that path:
- Keep backup servers and repositories off the production domain. Use local accounts or a separate management identity.
- Give backup administrators dedicated accounts that are never used for email or browsing.
- Require phishing-resistant MFA (FIDO2 security keys or platform passkeys) on the backup console and the cloud storage account.
- Store break-glass credentials offline, sealed, with every use logged.
Step 4: Gate destructive actions
If your platform supports multi-person approval for deleting jobs, removing restore points or changing retention, turn it on. Veeam, Rubrik, Cohesity and Commvault each offer a version of this control under different names. Whether you have it can depend on your software version and license tier, so check. Then set alerts on any change to retention or encryption settings.
Evidence to keep: screenshots of MFA enforcement, a list of backup admin accounts showing they aren't domain-joined, and the multi-person approval configuration.
Configure Object Lock Retention an Attacker Can't Shorten
Step 5: Set immutability at the storage layer
Underwriters recognize object lock more readily than any other backup control. Amazon S3 Object Lock, Azure immutable blob storage, Wasabi, Backblaze B2 and most S3-compatible on-premises appliances all support write-once retention. S3 has two modes. In governance mode, privileged users can override retention. In compliance mode, nobody can shorten or remove the lock before it expires, and that includes the root account. Use compliance mode, or your platform's locked-policy equivalent, for any copy you describe to an insurer. Azure, for example, separates unlocked from locked time-based retention policies. A lock that an admin can lift is a weak answer to the form's question.
Picking the retention window is a trade-off between protection and storage cost. Attackers often sit in a network for days or weeks before they encrypt anything, so a 7-day lock can run out on the last clean restore points you had. For most mid-size companies, 30 days of immutable retention on Tier 1 data is a defensible baseline. Regulated firms, and companies without round-the-clock monitoring, should consider 60 to 90 days.
Hypothetical example: 20 TB of protected data with daily incrementals held under a 30-day lock typically ends up as 35–45 TB stored. The exact figure depends on how fast your data changes. Typical 2026 cloud object storage rates run roughly $6–$25 per TB per month, depending on provider, storage tier and region. At those rates the bill is about $210–$1,100 a month before egress and API charges. These are indicative market ranges that vary by contract, so confirm pricing with vendors.
Evidence to keep: a screenshot of the bucket or container lock settings showing the mode and retention period, plus the configuration of the backup job that writes to it.
Choose an Isolation Layer That Matches Your Risk
Step 6: Add a copy outside your main blast radius
Immutability stops deletion. Isolation covers the case where the backup platform or the cloud account itself is compromised. Underwriters generally accept any of the options below as the isolated copy, provided you describe yours accurately.
| Option | Typical 2026 cost range (indicative) | Strength | Weak spot |
|---|---|---|---|
| Immutable cloud object storage in a separate account or tenant | $6–$25 per TB/month plus egress | Fast to deploy, available in most regions | Restore speed depends on bandwidth and egress fees |
| Hardened on-premises repository (for example, Linux with immutability flags) | $8,000–$40,000 hardware, plus admin time | Fast local restores, data stays in-country | You own physical security and OS hardening |
| Vendor-managed isolated vault or cyber-recovery service | Often $15–$60 per TB/month or a per-workload fee | Administrative control sits outside your own staff | Vendor lock-in; check where the data is held |
| Rotated offline tape or removable media | $5,000–$30,000 for drive/library, low media cost | A true physical air gap | Slow restores; only as good as rotation discipline |
For most companies under 1,000 staff, the better fit is a hardened local repository plus an immutable cloud copy in a separate account or tenant. The local copy gives you fast restores, the cloud copy gives you isolation, and together they usually cost less than a managed vault. Managed vaults earn their price when you lack in-house backup expertise or need administrative separation for a regulator. Data residency can narrow your choices. Sector rules in Saudi Arabia, the UAE or India, or a customer contract that requires in-country storage, may apply to you. If so, confirm the provider has a region that meets them before you sign.
Evidence to keep: a simple architecture diagram showing each copy, where it sits and which credentials can reach it.
Prove Restores Work and Capture the Proof
Step 7: Test on a fixed schedule
Until you've restored from a backup, you're assuming it works. A reasonable testing cadence:
- Monthly: automated restore verification of Tier 1 VMs or databases. Many platforms boot the restored image in an isolated sandbox and run health checks.
- Quarterly: a manual restore of a representative file share and one SaaS mailbox or site, pulled specifically from the immutable copy.
- Annually: a timed recovery of a full Tier 1 application, identity included, measured against your RTO.
Step 8: Record backup restore testing evidence properly
For each test, record:
- the date and the system restored
- the source copy (local, immutable cloud or vault)
- how old the restore point was
- how long recovery took
- the result, plus any problems found and how you fixed them
Keep the report the platform generates, not just a note in a ticket. If a test missed its RTO, record that too. An honest log that shows a failure and the fix is more credible than a spotless log nobody believes.
Financial entities in the EU should map this testing to DORA's requirements on backup policies and restoration testing. Several regulators in the Gulf, Singapore and India set similar expectations for the entities they supervise. Check the specific rulebook that applies to you.
Assemble the Underwriter Evidence Pack
Step 9: Package it before the application goes out
Clear, dated documentation cuts down the back-and-forth during underwriting and gives your broker something concrete to work with. Build a short pack, ten pages or fewer, that the broker can attach to the submission:
- A one-paragraph summary of your backup architecture, plus the diagram from Step 6.
- The tiered inventory and RPO/RTO table.
- Screenshots of the object lock mode and retention, MFA on backup consoles, and domain separation.
- The last three monthly restore reports and the most recent full recovery exercise.
- The retention policy, including how it squares with data-protection law.
That last item causes trouble across jurisdictions. The UK GDPR, the EU GDPR, Brazil's LGPD and India's DPDP Act all limit how long you may keep personal data. Long immutable retention needs a documented justification. You also need a process for honoring deletion requests once the locks expire. Have privacy counsel review the policy.
Be precise when you answer the form. "Daily backups of Tier 1 systems to S3 in compliance mode with 30-day retention, in a separate AWS account, restore-tested monthly" holds up far better than a bare "yes." If part of that isn't true yet, say so and give a completion date.
Keep the Setup Claim-Ready and Review It Each Renewal
Step 10: Preserve evidence for the claim, not just the application
During a live incident, the insurer's breach coach and forensic firm will ask for backup job logs, retention change logs and console access logs. Ship those logs to somewhere the attacker can't reach, such as a separate SIEM tenant or the immutable store itself. Keep them at least as long as your immutability window, and longer if a regulator requires it. India's CERT-In directions, for example, set a 180-day log retention requirement for many organizations. Keep a copy of the evidence pack outside your corporate environment too, so you can still reach it when email and file shares are down.
Fix these before renewal:
- Backup servers still joined to the production domain.
- Object lock in governance mode, or retention of 7 days or less.
- No restore taken from the immutable copy in the past quarter.
- SaaS data left out of backup scope.
- One person who can both change retention and approve that change.
The rule: if you can't produce a dated screenshot or report for a statement about your backups, either reword the answer or fix the control before you submit. Assemble the pack 60 days before renewal and have your broker review it. Then ask a licensed coverage advisor how your policy's warranty and misrepresentation clauses apply to the answers you give.
Questions buyers ask
Do insurers require a true physical air gap, or is immutable cloud storage enough?
Most underwriters treat logically isolated immutable storage as meeting the 'offline or immutable' question. A typical example is object lock in compliance mode, held in a separate account. Physical air gaps such as rotated tape are welcome, but mid-size companies are rarely required to have one. The test is whether stolen production credentials can delete or alter the protected copy. Wording varies by carrier, so confirm with your broker.
How long should immutable retention be for cyber insurance purposes?
There's no universal number, and immutable backup requirements vary by insurer and region. Thirty days on critical systems is a common, defensible baseline, because attackers often stay in a network for days or weeks before encrypting. Regulated firms often choose 60–90 days. Whatever you pick has to fit within data-protection limits on retention.
What counts as acceptable backup restore testing evidence?
The strongest evidence is a platform-generated restore report showing the date, system, source copy, age of the restore point, recovery time and result. Add notes on any failures and how they were fixed. A ticket that just says 'restore tested OK' carries little weight with underwriters or claims teams.
Does Microsoft 365 or Google Workspace data need its own immutable backup?
Usually, yes. Native SaaS retention features aren't built as independent, attacker-resistant backups, and someone who takes over an admin account can often change them. Third-party SaaS backup services from vendors such as Veeam, Druva, Keepit or AvePoint can write to immutable storage and close that gap.
What happens if I answered 'yes' to immutable backups but my setup doesn't fully match?
An inaccurate answer can give the insurer grounds to dispute or reduce a claim. How much depends on the policy wording and on local insurance law, which varies widely by country. Tell your broker straight away, correct the answer in writing and commit to a fix date. Have a licensed coverage advisor review how your policy treats misstatements.
Written by the Auditvault Pro editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.