By Prakhar Rai, Founder, My Legal Pal. Advocate, Bar Council of India. LL.B, NLSIU Bangalore. Master of Business Laws.
Quick answer: A SaaS security addendum is a schedule to your main agreement that sets the security measures you commit to. Promise the programme and the standard, not perfect outcomes. Put every security promise you actually mean in the addendum, state which document wins in a conflict, and say that questionnaire answers are information and not contract terms. Then check that your real controls match every sentence before you sign.
Key takeaways
- Your security questionnaire answers can come back to you as claims in a dispute. Decide in advance how they relate to the contract.
- Say which document ranks first when the MSA, the DPA and the addendum disagree. Silence leaves it to a judge.
- Promise to “maintain a security programme” with named measures. Do not promise that no breach will ever happen.
- A certification promise should name the standard, the scope and the audit period, not just “SOC 2 compliant”.
- Breach notice should run from when you confirm an incident, not from when something odd first appears in a log.
- Give audit rights in steps: your reports first, then a questionnaire, then a limited audit only if the reports fail to answer the question.
A buyer’s security team sends you a spreadsheet of 200 questions. Sales fills it in over a weekend. A month later procurement sends a “security addendum” with the order form and asks for signatures by Friday. Nobody has checked whether the spreadsheet and the addendum agree, or whether the engineering team can actually do what both say.
This guide is written from the vendor’s side. It shows what to put in the addendum, where vendors over-commit, and fallback positions you can offer when a customer pushes. It sits alongside our guides to SaaS agreement drafting and the data processing agreement. We do not repeat the data protection clauses here. We cover the security layer on top of them.
A common pattern is a vendor whose addendum promised annual penetration tests by an independent firm and notice of any incident within 24 hours, while the security team ran tests every two years and had no on-call process for weekends. The addendum was signed first. The processes came later, after the first customer audit request.
What is a SaaS security addendum, and do you need one?
Yes, if you sell to business customers that handle sensitive data. It is a schedule or exhibit to your main agreement that lists the security measures you maintain, how you report incidents, and what the customer can check. Some call it a security exhibit, an information security schedule or a technical and organisational measures annex.
Large customers will ask for one. If you do not offer your own, they will send theirs, and customer paper is usually written to suit the customer. Offering your own standard addendum early keeps the promises in line with what you actually do.
| Document | What it covers | Where security shows up |
|---|---|---|
| Master agreement (MSA) | Services, fees, term, liability | A short security clause and the liability cap |
| Data processing agreement (DPA) | Personal data rules under privacy law | Duty to keep personal data secure, breach notice, audits |
| Security addendum | Specific measures for the whole service, including non-personal data | Controls, certifications, testing, incident process |
| Questionnaire or trust page | Answers given before signing | Statements that may later be treated as promises |
Do your security questionnaire answers become contract terms?
They can, and you should not leave it to chance. A questionnaire answer is a statement you made to win the deal. Four routes can turn it into a problem.
- Incorporation. The contract refers to your questionnaire, response document or proposal, and so pulls the answers in as terms.
- Misrepresentation. If a customer relied on a false statement when it signed, it may claim damages. In England and Wales, section 2(1) of the Misrepresentation Act 1967 allows damages for a misrepresentation that induced a contract, even if it was not made fraudulently. Other countries have their own versions.
- Weak entire agreement clauses. A standard “entire agreement” clause may not stop a claim based on a statement made before signing. In England and Wales, section 3 of the same Act makes a clause that excludes liability for misrepresentation ineffective unless it passes a reasonableness test, and the party relying on it must show it is reasonable.
- Regulators. In the United States, the Federal Trade Commission treats false security claims as deceptive practices under section 5(a) of the FTC Act. In a 2020 action the Commission alleged that a video conferencing company told users its meetings were protected by end-to-end encryption and a 256-bit key, when its servers held the keys and it used a 128-bit key. The point for vendors is general: public security statements are checked against what you actually do.
The fix is a short set of habits and one clause.
| Position | Sample wording |
|---|---|
| Vendor favourable | “Security questionnaires, trust centre content and sales materials are provided for information only, reflect the Supplier’s practices at the date given, and are not terms of this agreement. The Supplier’s security commitments are those in the Security Addendum.” |
| Balanced | “The Supplier’s written responses to the Customer’s security questionnaire dated [date] are incorporated as a statement of its practices at that date. The Supplier will notify the Customer of any material reduction in those practices. The Security Addendum prevails if there is a conflict.” |
| Customer favourable | “The Supplier warrants that its responses to the Customer’s security questionnaire are accurate and complete and will remain so throughout the term.” |
Use the vendor wording as your opening position and the balanced one as your fallback. Never accept the customer wording without checking every answer first, because it turns a weekend spreadsheet into a continuing warranty.
Habits that protect you:
- Have security or engineering review every answer, not just sales.
- Date and version your standard answers, and keep a record of who approved them.
- Answer “no” or “partly” when that is true. A true “partly” costs less than a false “yes”.
- Keep your public trust page in line with the addendum.
Which document wins if the MSA, the DPA and the addendum disagree?
Whichever one your precedence clause ranks first, so write one. Without it, the customer will argue for the document that favours it, and so will you.
A common order is the DPA for personal data matters, then the security addendum, then the MSA. Most vendors prefer the MSA to rank higher, so that the liability cap and the exclusions apply to security promises too.
| Position | Sample wording |
|---|---|
| Vendor favourable | “If there is a conflict, this order applies: (1) the Order Form, (2) the Master Agreement, (3) the Security Addendum, (4) the DPA, except that the DPA prevails on the processing of personal data.” |
| Balanced | “The DPA prevails on personal data. The Security Addendum prevails over the Master Agreement on security measures. In all other cases the Master Agreement prevails.” |
| Customer favourable | “The Security Addendum and the DPA prevail over the Master Agreement and any Order Form.” |
Two checks before you sign. First, make sure the liability cap clause says whether security and data breaches are inside the cap or in a separate “super cap”. Second, make sure a customer’s own purchase order terms cannot override your addendum. Our post on the limitation of liability clause explains how caps work.
Should the addendum promise outcomes or promise a security programme?
Promise the programme. No vendor can promise that a breach will never happen. A clause that says you “will protect” data can be read as a promise of the result, and a breach then looks like a breach of contract.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier will maintain a written information security programme with administrative, technical and physical safeguards that are appropriate to the risk and no less protective than described in this Addendum.” |
| Balanced | “The Supplier will implement and maintain the measures in Schedule 1 and will review them at least once a year against [ISO/IEC 27001 or the NIST Cybersecurity Framework].” |
| Customer favourable | “The Supplier will protect Customer Data against all unauthorised access, loss or disclosure.” |
The vendor and balanced wording follows how privacy law is already written. Article 32 of the GDPR asks controllers and processors to implement measures “appropriate to the risk”, taking into account the state of the art, the costs of implementation and the nature, scope, context and purposes of processing. That is a standard of reasonable effort, not a guarantee. Many other laws use similar language, so this is a position customers recognise.
Can you promise to follow ISO 27001, SOC 2 or NIST?
Yes, but be exact about what you are promising. A framework name alone is vague. Each standard means something different, and customers read them in their own favour.
| Standard | What it is | Safe wording | Avoid |
|---|---|---|---|
| ISO/IEC 27001 | A standard for an information security management system, with certification by an accredited body | “Maintain ISO/IEC 27001 certification for the scope covering the Service” | “ISO compliant” with no scope |
| SOC 2 | An audit report by an independent auditor against the AICPA trust services criteria: security, availability, processing integrity, confidentiality and privacy | “Provide the most recent SOC 2 Type 2 report on request, under confidentiality” | “SOC 2 certified”. It is a report, not a certificate |
| NIST Cybersecurity Framework 2.0 | A voluntary framework, released in February 2024, organised around six functions: govern, identify, protect, detect, respond and recover | “Align its programme with the NIST Cybersecurity Framework” | “Fully implements NIST” |
Three drafting points:
- Type 1 versus Type 2. A Type 1 report looks at the design of controls at one date. A Type 2 report covers a period and tests whether the controls worked. Do not promise Type 2 if you only have Type 1.
- Scope. State which products and sites the report or certificate covers. Many reports cover one product, not the whole company.
- Loss of certification. Decide in advance what happens if a certificate lapses. A short cure period and a duty to tell the customer is fairer than an automatic right to terminate.
Under GDPR Article 32(3), adherence to an approved certification mechanism may be used as an element to demonstrate compliance with the security duty. That helps, but only for certification schemes approved under that Regulation. It does not make a general ISO or SOC 2 claim count automatically.
How specific should the technical controls be?
Specific enough to be useful, general enough that you can change them. The usual vendor mistake is writing down today’s tool or setting as a permanent promise.
| Topic | Over-committing | Safer wording |
|---|---|---|
| Encryption | “AES-256 on all data at all times” | “Encrypt Customer Data in transit and at rest using industry standard methods” |
| Access | “Multi-factor authentication for every user” | “Use multi-factor authentication for administrative access to production systems” |
| Logging | “Retain all logs for 7 years” | “Keep security logs for a period that supports incident investigation, at least [x] months” |
| Patching | “Fix all vulnerabilities within 24 hours” | “Remediate critical vulnerabilities within a timeline set by risk, and critical ones within [x] days” |
| Hosting | “Data stays in one named data centre” | “Host in [region] with the providers listed in the subprocessor list” |
| Staff | “Background checks on all staff” | “Carry out background checks on staff with production access, where law permits” |
Add this line, and customers usually accept it: “The Supplier may update its measures at any time, provided the update does not materially reduce the overall level of security.”
Can you promise to report a security incident within a fixed number of hours?
Yes, if the clock starts when you confirm an incident and the notice is tied to your real process. Customers who are controllers under privacy law often have a tight deadline of their own. Under Article 33 of the GDPR a controller must notify the supervisory authority within 72 hours of becoming aware of a personal data breach, where feasible, and a processor must notify the controller “without undue delay” after becoming aware. That is why customers ask for 24 or 48 hours. Our guide to the first 72 hours after a data breach covers the customer side of that clock.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier will notify the Customer without undue delay, and in any event within [72] hours, after confirming a Security Incident affecting Customer Data.” |
| Balanced | “The Supplier will notify the Customer without undue delay, and within [48] hours, after becoming aware of a Security Incident affecting Customer Data, and will give updates as new information becomes available.” |
| Customer favourable | “The Supplier will notify the Customer within [24] hours of any actual or suspected unauthorised access to Customer Data.” |
Points to settle:
- Define “Security Incident”. Limit it to confirmed unauthorised access to, or loss of, Customer Data. Exclude failed login attempts, port scans and blocked attacks.
- Say what the notice contains. What you know at the time, with details to follow. Do not promise a full root cause report in the first message.
- Check your process. Do you have someone on call at the weekend who can decide that an incident is confirmed? If not, do not promise 24 hours.
- Say who pays. Costs of notifying individuals, credit monitoring and regulator fines belong in the liability clause, not in the incident clause.
How much audit access should you give a customer?
Give it in steps, and start with documents. Letting every customer walk into your engineering team does not scale and can put other customers’ data at risk. Privacy law does expect some audit rights for processors. Article 28(3)(h) of the GDPR requires a processor to make available the information needed to show compliance, and to allow for and contribute to audits, including inspections, by the controller or another auditor it appoints. A step process meets that and protects you.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier will give the Customer its current audit report or certificate and answer reasonable written questions once a year. An on-site audit is available only if these do not answer the Customer’s question, or after a Security Incident.” |
| Balanced | “The Customer may audit once in any 12 months on [30] days’ notice, during business hours, under confidentiality, at the Customer’s cost, and without access to other customers’ data. The Supplier’s current audit report satisfies the audit request where it covers the same scope.” |
| Customer favourable | “The Customer may audit the Supplier’s systems, facilities and records at any time.” |
Add these limits: a named auditor who is not a competitor, a signed confidentiality agreement, a duty to give the findings to you, and a rule that the customer pays for the time you spend if the audit takes more than [x] days.
Should you promise penetration tests and share the results?
Promise the test, share a summary, and keep the report. A full penetration test report is a map of your weak points. Handing it to every customer creates a risk of its own.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier will have an independent third party carry out penetration testing of the Service at least once a year and will give the Customer a summary of the results on request, under confidentiality.” |
| Balanced | “The Supplier will give the Customer an executive summary of the latest penetration test and confirm that critical and high findings have been fixed or have a remediation plan.” |
| Customer favourable | “The Supplier will give the Customer the full penetration test report and allow the Customer to run its own tests.” |
Customer-run tests need a separate rule. Allow them only on a staging environment, with a written scope, advance notice, and a duty to report findings to you first. A test run against production can look like an attack, and it can affect other customers.
What security duties should pass down to subprocessors?
The same level of protection, in writing, with you staying responsible. Customers will ask about your cloud host, support tools and any AI provider you use. Your addendum should say how you choose and monitor them. Our SaaS IP clauses guide explains why the vendor chain matters for customer data and AI features.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier will carry out security due diligence on subprocessors before use and will bind them to written security obligations that are no less protective in substance than this Addendum.” |
| Balanced | “The Supplier will bind subprocessors that access Customer Data to written security terms and remains responsible for their performance of those terms, subject to the limits in the Master Agreement.” |
| Customer favourable | “The Supplier will be liable for all acts and omissions of its subprocessors as if they were its own, without limit.” |
One warning. If your own cloud provider gives you weaker terms than you give customers, you carry the gap. We cover that mismatch in our post on cloud outages and force majeure.
Can you change the security addendum after signing?
Yes, if you reserve the right and give a limit. Security improves over time, and you do not want to be bound to a 2026 setting in 2030.
| Position | Sample wording |
|---|---|
| Vendor favourable | “The Supplier may update this Addendum and its security measures from time to time, provided the overall level of security of the Service is not materially reduced.” |
| Balanced | “The Supplier will give [30] days’ notice of any change that reduces a measure in Schedule 1. If the reduction is material, the Customer may end the affected Services and receive a refund of prepaid fees for the unused period.” |
| Customer favourable | “No change to this Addendum has effect unless the Customer agrees to it in writing.” |
Mistakes vendors make with security addenda
| Mistake | Why it hurts | Fix |
|---|---|---|
| Signing the customer’s addendum without comparing it with your questionnaire answers | The two can contradict each other | Reconcile both before signature |
| Sales fills in the questionnaire alone | Wrong answers become claims | Security reviews and signs off every answer |
| Promising “SOC 2 certified” | It is a report, not a certificate | Describe it as a report, with the period and scope |
| No precedence clause | Argument over which document applies | Rank the documents in the contract |
| Broad definition of “Security Incident” | Every scan becomes a reportable event | Limit it to confirmed unauthorised access to Customer Data |
| Fixed hour deadline with no on-call process | You miss it on a Saturday | Promise only what the team can do |
| Unlimited audit rights | Cost and risk to other customers | Reports first, then a limited audit |
| No right to update the measures | You are stuck with old controls | Reserve the right, with a no-material-reduction limit |
| Security failures outside the liability cap discussion | Surprise uncapped exposure | Decide whether a super cap applies |
A checklist before you sign a security addendum
- Have engineering and security read every sentence, not only legal?
- Does each promise match what you do today, and can you keep doing it?
- Have you compared the addendum with all questionnaire answers and your trust page?
- Is there a clause saying questionnaire answers are information and not terms?
- Is the order of precedence between the MSA, DPA and addendum stated?
- Does the addendum promise a programme and standard, not a result?
- Are certifications named with their scope, period and type of report?
- Does the incident definition exclude failed and blocked attempts?
- Does the notice clock start at confirmation, and does your on-call process match it?
- Are audit rights in steps, with confidentiality and cost rules?
- Are subprocessor duties tied to written terms you can really get from them?
- Can you update the measures without lowering the overall level of security?
- Is the security liability position clear in the cap clause?
- Is someone responsible for checking compliance each year?
Get your security addendum written once and used many times
A standard addendum, a short set of approved questionnaire answers and a clear precedence clause save weeks on every enterprise deal. If you want a security addendum drafted or an incoming customer addendum reviewed, we can review your agreement for your jurisdiction. Start with our SaaS agreement template, see the legal documents every SaaS startup needs, or speak to our technology lawyers for tech startups, SaaS and AI platforms.
Quick answers
Is a security addendum the same as a DPA? No. A DPA deals with personal data under privacy law. A security addendum covers the security measures for the whole service, including data that is not personal data. Many deals need both.
Do security questionnaire answers form part of the contract? Only if the contract incorporates them or a claim is built on them, but they can still lead to a misrepresentation claim. Say in the contract that they are information, and keep them accurate.
Does the DPA or the security addendum take priority? Whichever your precedence clause says. A common order puts the DPA first for personal data and the addendum first for security measures.
Can a SaaS vendor refuse to give a customer its penetration test report? Yes, in most deals. Offer an executive summary and confirmation that serious findings are fixed or planned.
Is SOC 2 a certification? No. It is an independent auditor’s report against the trust services criteria. Describe it as a report, and name the period and scope.
Should a vendor promise to notify a breach within 24 hours? Only if the team can really do it. Many vendors offer 48 or 72 hours from confirmation of an incident, with updates after.
Does GDPR require processors to allow audits? Yes. Article 28(3)(h) requires a processor to make information available and to allow for and contribute to audits, including inspections. The contract can set a sensible process.
Can I change the security measures after signing? Yes, if the contract lets you, and you do not materially reduce the overall level of security. Customers often ask for notice of reductions.
Does the security addendum sit inside the liability cap? That depends on what the contract says. Many customers ask for a higher cap or a separate cap for data breaches, so decide your position before the negotiation starts.
A note on who we are
My Legal Pal is an international team of contract lawyers. This article is general information, not legal advice on your facts. We can review your agreement for your jurisdiction.
About the author. Prakhar Rai is an advocate enrolled with the Bar Council of India and the founder of My Legal Pal. He holds an LL.B from NLSIU Bangalore and a Master of Business Laws, and works on contract drafting, review and disputes for founders and businesses across borders. Full profile


