Addressing information security within supplier agreements
Annex A 5.20 turns supplier security from a promise into a signed agreement, with terms matched to each supplier's tier and risk. A tiered security schedule and a current supplier register give auditors and regulated customers evidence rather than assurances. Contract negotiation stays human, but the register, onboarding gate and assurance chasing can be automated.
The due-diligence questionnaire asks for the security schedule in your critical supplier's contract. You find a purchase order and an NDA. This guide covers ISO 27001 Annex A 5.20, the control that turns supplier security from a promise into a signed, risk-matched agreement, for security and compliance leads at regulated service providers and B2B SaaS firms.
You will get the control in plain English, the clauses that carry weight, a route through the AWS and Microsoft standard-terms problem, the register auditors sample from, and the questions asked in the room. It is not a mountain. Most of it is a template set, a register and one gated onboarding step.
ISO 27001 Annex A 5.20 requires the security expectations you place on each supplier to be written into a binding agreement, matched to the relationship's type and risk. Both parties' obligations must be explicit, and the agreements need recording in a register and revisiting so terms stay fit for what the supplier now does.
A supplier's promise to look after your data is worth what the contract says it is worth. A payroll processor and a stationery vendor do not warrant the same terms, so the control asks you to match each agreement to the relationship.
In practice, the control asks for three things:
Annex A 5.20 is a preventive control that covers confidentiality, integrity and availability. Notification duties, handling rules and exit obligations are agreed before data flows, so failures are limited in advance rather than discovered in the wreckage.
For a 50-person firm, that means a small set of templates and a register, not a legal programme.
In the supplier-control chain, A.5.19 sets the policy and tiering, A.5.20 converts that policy into enforceable contract terms, A.5.21 extends the reach down the ICT supply chain, and A.5.22 checks supplier performance against what was signed. Monitoring can only measure what the contract puts in writing.
Auditors read the policy under A.5.19 and the contracts under A.5.20. Your tiering decides which schedule a supplier receives, so vague tiers produce contracts that are either over-papered or too thin to matter.
Each control leaves a different trace in your evidence:
The sub-processor and flow-down clauses in your schedule are where A.5.20 hands over to the ICT supply chain. They bind the supplier's own suppliers to equivalent terms. A review meeting that checks a supplier against a clause nobody agreed has nothing to enforce, so work in order: tier, contract, monitoring.
A supplier security schedule covers what information the supplier receives and how it is handled, applicable legal duties, who may access your assets, a hard incident-notification window, sub-processing rules, usable assurance, continuity expectations, and exit terms. Build one per supplier tier and attach it at onboarding, so legal input happens once.
For anything touching your data, these are the clauses that auditors and incidents test:
Re-express these in your own words rather than copying anyone's list, but address each one for critical suppliers.
The notification window is the clause that bites first. The ICO personal data breach reporting guidance says a notifiable breach must reach the ICO without undue delay and within 72 hours. A supplier clause that repeats "without undue delay" gives you nothing to enforce while that deadline runs, so name the hours and require logs and cooperation.
Flow-down deserves the same precision. The NCSC supply chain security principles advise putting your minimum security requirements in supplier contracts and requiring suppliers to pass them to their own sub-contractors. Add a notice right on top, so you hear before a sub-processor changes, not after.
Four pitfalls recur, and each clause above inverts one:
An NDA is a secrecy agreement, not a supplier security control.
You do not redline hyperscalers. Review their standard terms and security addenda, collect their independent assurance, and record any residual gap in your risk register with a named owner. That documented judgement, not a bespoke clause, is what an auditor wants to see for click-through and non-negotiable SaaS terms.
Work through three steps:
The register entry then points to the assurance file and the risk entry, so an auditor can follow your judgement in one pass. Sometimes the gap is too wide to accept, and the documented decision is to choose another supplier or add your own controls.
Proportion cuts the other way too. Demanding bespoke audit rights from a hyperscaler buys risk theatre, not protection. Putting the full critical-tier schedule on a trivial vendor is the same mistake. Low-risk vendors that never touch confidential data may need only confidentiality basics. Over-papering the stationery supplier is wasted legal spend and a review burden you will carry forever.
An ISO 27001 supplier register lists every agreement with its risk tier, contract status, security-terms version, data processing agreement status, assurance validity and renewal date. Auditors sample from it, so it must be current within the last review cycle and show that each contract is still needed and still fits.
Add the signature date to each entry, and schedule a periodic pass to confirm every contract still matches what the supplier delivers. Contracts signed once and never revisited are a common failure: the supplier's role has tripled in scope while the paper still describes the original job.
Version your tier templates and record the version against each supplier. When a template changes, after a new regulation or a lesson from an incident, the register shows at once which suppliers still sit on older terms. That list becomes your re-papering plan.
The register is also the index to the rest of your evidence. For each sampled supplier, expect to produce:
Check scope when you file the assurance, not when the auditor asks. Accepting whatever certificate a supplier sends, without reading its date or scope, is a quiet failure that surfaces in the audit room.
A certificate that covers a supplier's head office says nothing about the service you actually consume.
For ISO 27001 A.5.20, auditors usually pick two or three suppliers from your register, most often the most critical plus one recent addition, and read the signed terms against your policy. They then check the promised assurance is on file and in scope, and trace recently terminated suppliers through their exit clauses.
Expect questions like these in the room:
Run the sample yourself before the audit: your most critical supplier, one added in the last year and one that has left. If any of the three cannot be traced from register to signed terms to assurance on file, you have found the finding before the auditor did.
Certification bodies typically grade by severity and pattern:
A current register and a tiered template set make all three traces quick to answer.
Contract negotiation stays human, but the record-keeping around it automates well. Hold the register as structured data, gate onboarding so no supplier goes live without signed tier terms, chase assurance against expiry dates, and watch sub-processor pages for changes. The control then becomes impossible to skip, rather than dependent on memory.
Each piece works like this:
A supplier register nobody can query is an inventory, not evidence.
Defining what each tier's terms must contain, and making the accept-or-reject call on non-negotiable vendor terms, is Fractional CISO work. The gated workflow, register-as-data and assurance chasing are GRC engineering deliverables.
The table compares a manually kept register with an engineered one.
| Evidence question | Manual register | Engineered register |
|---|---|---|
| Which critical suppliers lack current terms? | Filter a spreadsheet, if someone updated it | Query on structured records |
| Renewal dates | Calendar reminders | Ticket raised ahead of expiry |
| Supplier go-live | Procurement remembers to ask | Workflow blocks until tier terms are signed |
| Assurance expiry | Chased when noticed | Automated request, filed against the entry |
| Sub-processor changes | Found at the next review | Page watch raises a signal |
A firm serving regulated clients sits downstream of its customers' obligations. DORA and NIS2 both push contractual requirements through the supply chain, and enterprise procurement teams increasingly ask to see supplier terms, not just policy. Signed, tiered agreements therefore answer diligence questions with evidence instead of promises.
The gap usually surfaces at the worst moment. A diligence exercise or a live incident reveals that a critical vendor was onboarded on a bare order form, with no logs to demand, no duty to cooperate and no recourse.
The pack a regulated buyer wants is the one you already keep for the auditor: the register extract, the tiered schedule template and the assurance on file for your critical suppliers. Share the template and the register extract rather than signed contracts, which usually carry their own confidentiality terms.
At 50 people, a tiered schedule set and a current register let you answer a regulated buyer's diligence from files you already hold.
No. The control cares that the terms are binding and present, not where they sit. They can live in the master agreement, a schedule or an addendum. A separate schedule per tier is simply easier to reuse across suppliers and to update at renewal.
No. A DPA belongs wherever the supplier processes personal data on your behalf, and a stationery vendor typically does not. The two documents do different jobs: the DPA sets data protection terms for personal data, while the security schedule protects all the information the supplier touches.
Whoever owns contracts: legal, procurement or, in a small firm, the COO or founder. Security defines what each tier's schedule must contain. The control fails when security assumes legal has it and legal assumes the reverse, so name one owner in writing.
Prioritise by tier, with critical suppliers first. Fix them at renewal where possible, and by addendum where the gap is urgent. Auditors accept a credible remediation plan with dates; they do not accept a gap that nobody noticed.
Templates and a register typically take two to four weeks with legal input. Fixing existing critical suppliers is the long tail, because it depends on each supplier's renewal date and signing speed. Run it top tier first, so the highest risk closes earliest.
Every control in this framework and the clause guides that sit above them, grouped by control domain. Open any entry for its control definition, implementation steps and the audit evidence it expects. Dimmed entries are planned.
Work with a Trust Architect to scope your ISMS, close the gaps, and get audit-ready — without stalling your roadmap.