Security at Mantle

You're trusting us with the most sensitive data your business holds.

Here is how we earn it. In detail, in plain language, and verifiable where it matters. Mantle holds the bookkeeping, banking, payroll, and tax records of thousands of businesses across Europe. The same standard of care your clients expect of you, we apply to ours.

Read our Privacy Policy

Reach the team directly at [email protected].

GDPR compliant

EU controller with full data subject rights. Standard contractual clauses where they apply, no transfers outside the EEA by default.

EU data residency

All customer data is stored and processed in the European Union, across providers whose data centres carry ISO 27001 certification.

PSD2 via GoCardless

Bank connectivity through GoCardless, an FCA-authorised AISP and PISP. We never see or store your bank credentials.

How we think about security

Four ideas that shape every decision

We don't have every certification yet, and we say so plainly. What follows is the work we do today, expressed as the principles that produced it.

Defense in depth

Multiple independent layers across encryption, identity, network, and application. No single failure exposes data, and no single component is treated as trusted.

Least privilege

Internally, every employee. In the product, every user. In our AI agent, every action. Each is granted the minimum access required for its purpose, and no more.

Transparency

We publish our subprocessors, our practices, our status, and our incident history. You can verify what we say, not only trust it.

Customer agency

You can see who has access, what they did, export your data, and delete it. Your security posture is yours to own, and we give you the tools to own it.

Architecture

Defense in depth, from disk to delivery

The controls below span encryption, network, isolation, and application layers. Each is implemented today. Ask us about any of them in detail.

Encryption & key management

AES-256 at rest
All stored data encrypted with AES-256. Keys managed by a cloud KMS, with HSM-backed protection for critical key material.
TLS 1.3 in transit
External traffic over TLS 1.3 with modern cipher suites. Mutual TLS between internal services.
Key separation & rotation
Encryption keys logically separated from customer data. Rotated annually, with envelope encryption for sensitive fields.

Network & isolation

Strict tenant isolation
Each business is isolated at the application and database layer. Cross-tenant access is impossible by design.
EU multi-cloud
Hosted on multiple cloud providers across EU regions. Their data centres carry ISO 27001 and ISO 27017/18 certifications.
Private network
Private VPCs, segmented subnets, no public database endpoints. Edge WAF and DDoS protection at the network edge.
Production isolation
Production data is never present in staging or test. Pre-production environments use synthetic or anonymised fixtures.

Application defences

Secure SDLC
Mandatory peer review, static analysis, dependency scanning, and threat modelling for new surface area.
OWASP-aligned
Top 10 controls baked into development standards. OWASP LLM Top 10 applied to every AI feature.
Continuous scanning
Severity-based remediation SLAs. Critical issues addressed within 7 days, high within 30 days.
Annual penetration tests
Third-party authenticated and unauthenticated tests at least annually. Executive summary available under NDA.
Web Application Firewall
Edge WAF blocks common attacks including injection, cross-site scripting, and automated abuse.
Responsible disclosure
Coordinated disclosure policy and security.txt published. Report a vulnerability to [email protected].

What you control

The toolkit you own

Security in Mantle is a partnership. You decide who sees what, observe what they do, and step in when something looks off. The capabilities below are available to your admins from day one.

Multi-factor authentication.
Available for every user and enforceable organisation-wide. TOTP authenticator apps and WebAuthn or FIDO2 security keys.
Single sign-on (SAML 2.0).
On Enterprise. Connects to Okta, Microsoft Entra ID, Google Workspace, and others through your existing identity provider.
Role-based access control.
Granular permissions per module and per action. Permission changes are recorded and reviewable by your admins.
Customer-visible audit logs.
Every privileged action visible to your admins with who, what, when, and from where. Exportable to your SIEM or as a file.
Session controls.
Idle timeout, per-device session revocation, and sign-out-of-all-devices in one click.
Modern password policy.
NIST SP 800-63B aligned. Length over complexity, no forced rotation, checks against breached-password lists.
IP allowlisting.
Enterprise tier. Restrict access to specific IP ranges or corporate networks.

Workspace access

Acme Holdings Ltd, 24 members

Every permission change is audit-logged

Just-in-time access request

Reviewed and bounded in time

Awaiting approval

Duration

45 minutes

Scope

Read-only, tenant scoped

FIDO2 verified

Yes

Reason

Customer ticket #4823. Investigating an invoice that fails reconciliation.

Approval, access, and revocation all enter the audit trail

What we control internally

No standing access. Every action accounted for.

Our engineers do not have always-on access to customer data. When access is required, it is requested, approved, time-bound, and logged. The discipline below is what stops insider risk from being theoretical.

Just-in-time production access.
Employee access is granted by need, approved by a second engineer, time-bound, and fully logged. No standing access to customer data.
Phishing-resistant employee MFA.
FIDO2 security keys for every engineer and operator with production access. No SMS, no recovery codes, no shared secrets.
Background checks.
Pre-employment screening for all staff who could come into contact with customer data.
Centrally managed devices.
Company laptops enrolled in MDM, full-disk encryption enforced, endpoint detection and response running on every machine.
Zero Trust network.
Every access request is authenticated and authorised regardless of network origin. No trusted internal network, no VPN bypass.
Joiner, mover, leaver discipline.
Identity and access provisioned and deprovisioned through central identity. Access revoked the day someone leaves.

Detect and respond

When something goes wrong, you hear about it

The fastest way to lose trust is to leave customers in the dark. Our detection and response practice exists to catch problems early, contain them quickly, and tell you within hours, not days.

Centralised logging.
Application and infrastructure logs aggregated and correlated with behavioural events.
Continuous monitoring.
Authentication, access, and infrastructure-level signals monitored continuously, with an on-call rotation for security alerts.
Intrusion detection and prevention.
IDS and IPS at the network layer to detect and block unauthorised access attempts in real time.
Tested incident response playbooks.
Documented playbooks rehearsed at least annually. Named incident commanders for each severity level.
48-hour breach notification SLA.
GDPR requires 72 hours to authorities. We commit to notifying affected customers within 48 hours of a confirmed breach.
Public status page.
Live availability and incident history. Subscribe to be notified of incidents the moment they are declared.

Incident response timeline

From signal to customer notification

  • Signal detected

    T+0

    Anomalous authentication pattern flagged by continuous monitoring.

  • Triage and contain

    T+15m

    On-call engineer paged. Containment actions executed under runbook.

  • Investigation

    T+2h

    Scope determined. Affected tenants and data classes identified.

  • Customer notification

    T+48h

    Affected customers contacted with scope, impact, and next steps.

48-hour notification commitment

GDPR allows 72 hours to authorities. We commit to 48 hours to you.

Vector action log

All agent activity, attributable to a user

Live
  • Drafted purchase invoice GH-2384

    Initiated by [email protected], awaiting approval

    14:02

  • Reconciled 17 bank lines

    BoC current account, March 1 to March 15

    13:58

  • Suggested supplier match

    A.G. Leventis Ltd, mapped from PO-1284

    13:51

  • Awaiting approval

    Close fiscal period February 2026

    13:44

Privileged actions require explicit human approval

Vector & AI security

Autonomous in capability, accountable by design

Vector is the AI agent that operates inside Mantle. It is more capable than any human user, so the constraints we place on it are stricter.

Bounded by your permissions.
Vector acts within the same role-based envelope as the requesting user. It cannot escalate privileges or access data the user could not already see.
Your data is never used to train models.
Customer data is not used to train, fine-tune, or improve any foundation model. First- and thirty-party model providers operate under zero-retention enterprise terms.
Human approval for high-impact actions.
Posting transactions, closing periods, executing payments, and sending external messages all require explicit human approval before they happen.
Full action audit trail.
Every Vector action is logged with the originating user, the prompt, the model version, the tools called, and the outcome, visible to your admins.
Prompt injection defences.
Inputs are sandboxed, tool calls are scoped to the user's privileges, and outputs pass through guardrails before any privileged action executes.
AI subprocessors disclosed.
We use OpenAI and Anthropic under enterprise contracts, listed publicly on our subprocessor page alongside their roles and jurisdictions.

Privacy & data handling

Your data is yours, and we make that obvious

EU-only residency, a published subprocessor list, plain-language descriptions of what we collect and why. The mechanics of GDPR, made operational and self-service wherever they can be.

EU-only by default.
All customer data is stored and processed within the European Union. SCCs are in place for any out-of-EEA transfer, which is not our default.
Subprocessor transparency.
A public, versioned list of every third party with potential access to customer data, with their purpose and jurisdiction.
30-day retention after termination.
Customer data is deleted within 30 days of contract end unless legal retention obligations require otherwise.
Customer-initiated export and deletion.
Your admins can export your data at any time and delete it under GDPR Article 17 right to erasure.
Named privacy contact.
Reach our privacy team at [email protected] for SARs, complaints, DPA requests, or questions about lawful basis.

Your data, your control

Self-service rights, available to every admin

  • Export your data

    CSV, Excel, JSON, accountant-compatible

  • Delete on request

    GDPR Article 17, in-app and on contract end

  • Subject access requests

    Standard SAR workflow, 30-day response

  • EU residency by default

    No transfers outside the EEA unless authorised

  • Subprocessor transparency

    Versioned public list with change notifications

EU-only storage and processing by default

Resilience & continuity

Built to keep going when it counts

Month-end. Payroll day. Year-close. An ERP outage is a business outage. These are the commitments we plan against, test against, and publish.

48h
Breach notification commitment
365 days
Encrypted backup retention
≤ 1h
Recovery point objective
≤ 4h
Recovery time objective

Backups are encrypted and geographically distributed. Disaster recovery is tested annually with documented results. Business continuity policy reviewed annually.

ERP-specific surfaces

Where the data is most sensitive

An ERP touches money, identity, and government filings. Each of these surfaces gets its own treatment, called out here so you know exactly how the riskiest flows are handled.

Bank feeds

We never see or store your bank password. GoCardless, an FCA-authorised AISP and PISP, handles consent and credentials end-to-end under PSD2.

Document recognition

Uploaded invoices and receipts are processed in our own infrastructure. The personal data of your customers and suppliers is held to the same standard as your own.

Payroll & employee data

National IDs, tax numbers, and salary figures sit in the strictest data class, with tighter access controls and field-level protections where they apply.

Tax filings

Submissions to government portals like TFA, IR, and VIES happen only under your explicit authorisation. We never transmit on your behalf without consent.

People, ethics, and conduct

Security depends on the people who build it

The behaviours and policies we hold ourselves to, alongside the controls applied to the people behind the platform.

People security

Background checks
Pre-employment screening for staff with potential access to customer data.
Joiner, mover, leaver
Identity and access provisioned and deprovisioned through central identity. Access revoked the day someone leaves.
Acceptable Use Policy
All employees and contractors acknowledge our internal AUP on joining.
Confidentiality
Every staff member and contractor operates under written confidentiality obligations.
Annual security training
Security awareness training for all staff, secure-coding training for engineers, and phishing simulations.
Third-party risk management
Vendor due diligence before onboarding, with ongoing review of suppliers with access to customer data.

Ethics & conduct

Code of Ethics
Public code applicable to all employees and contractors, governing day-to-day decisions.
Anti-bribery & corruption
Formal policy and annual training. Zero tolerance for facilitation payments or kickbacks.
Anti-modern-slavery
Statement aligned with the UK Modern Slavery Act and equivalent EU instruments.
Fair labour
Public statement of fair-labour practices across all employment and contractor relationships.
Whistleblowing programme
Confidential reporting channel with a formal non-retaliation policy.
Supplier Code of Conduct
The behaviours we require of every supplier, applied during onboarding and reviewed annually.

Trust artifacts

What to take to your procurement team

The documents and assessments you can request from us. Some are public. Others are available on request, or under NDA.

  • Privacy Policy

    How we collect, use, and process personal information. Linked from every page footer.

    Public
  • Service Level Agreement

    Availability commitments, support response times, and credits for missed targets.

    Public
  • security.txt

    Coordinated disclosure contact published at /.well-known/security.txt per RFC 9116.

    Public
  • Data Processing Addendum

    Standard DPA template aligned with GDPR. Signed by us on request.

    On request
  • Security overview one-pager

    Plain-language summary of our practices, suitable for sharing with stakeholders.

    On request
  • Vendor questionnaires

    We respond to CAIQ Lite, SIG Lite, and VSA-Q. Custom questionnaires considered on request.

    On request
  • Subprocessor list

    Every third party with potential access to customer data, versioned with change notifications.

    Under NDA
  • Architecture and data-flow diagrams

    Sanitised technical diagrams of our platform, available to evaluators under NDA.

    Under NDA
  • Penetration test summary

    Executive summary of our most recent third-party penetration test.

    Under NDA
  • PII evidence

    Certificate of coverage for our professional indemnity insurance.

    Under NDA

Need one of these? .

Frequently asked questions

The questions buyers, IT teams, and procurement raise the most. If yours is missing, send it to us and we will respond, then add it here.

Where is my data stored?
All customer data is stored and processed in the European Union, across providers whose data centres carry ISO 27001 and ISO 27017/18 certifications.
Who at Mantle can access my data?
Production access is granted by need, time-bound, and fully logged. No employee has standing access to customer data. Every privileged action by Mantle staff is recorded and reviewable.
Is my data used to train AI models?
No. Customer data is never used to train, fine-tune, or improve any foundation model. First- and thirty-party model providers operate under enterprise contracts with zero-retention and no-training terms.
Why don't you have ISO 27001 or SOC 2?
We are not yet certified. We have chosen to publish the controls themselves rather than rely on a badge while certification is in progress. We respond to vendor questionnaires (CAIQ Lite, SIG Lite, VSA-Q) and welcome detailed scrutiny of our practices.
Can I get a copy of your penetration test report?
We share an executive summary of our most recent third-party test under NDA. Contact our security team to start the process.
What happens to my data if I cancel?
Customer data is deleted within 30 days of contract termination, unless a legal retention obligation requires otherwise. You can export everything before that window closes.
What happens if Mantle is acquired or shuts down?
Our terms commit us to giving customers reasonable notice and a data export window in either scenario. The default is your data leaves with you in machine-readable formats.
What is your breach notification commitment?
We commit to notifying affected customers within 48 hours of a confirmed breach. GDPR requires 72 hours to authorities. We hold ourselves to a faster standard for the people whose data is involved.
Do you support SSO and MFA?
Yes. SAML 2.0 SSO on Enterprise. MFA with TOTP authenticator apps and WebAuthn or FIDO2 security keys is available for every user and can be enforced organisation-wide.
Are you used by regulated entities?
Yes. Mantle is used by accounting firms, investment managers, payment institutions, and other entities operating under European financial regulation. Customer references available under NDA.
Is Mantle suitable for processing health data?
No. Mantle is not designed for protected health information or other special-category health data. We ask customers not to upload medical records or similar content.
Can our auditor access the data they need?
Yes. Auditor and accountant access is a first-class feature in Mantle. Grant access by role, revoke when the engagement ends, and every action they take is in the audit log.
How do you handle subject access requests?
Send the request to [email protected]. We acknowledge within 5 business days and respond within 30 days, the GDPR statutory window.

Bring the hard questions

Our security team answers procurement questionnaires, walks evaluators through our architecture, and signs DPAs. The harder the question, the more useful the conversation.

Read our Privacy Policy

Or write to [email protected].