RBI Cyber Resilience Directions Bind Payment Operators. Six Clauses Land on Their Vendors.
The 5 November rulebook applies to authorised non-bank payment operators. Paragraph 5 and the audit scope carry it into every vendor contract they sign.
On 5 November 2026 the Reserve Bank of India published the Master Directions on Cyber Resilience and Digital Payment Security Controls for non-bank Payment System Operators (RBI/DPSS/2026-27/475). The RBI cyber resilience directions are addressed to the chairmen and chief executives of authorised operators: NPCI, the card networks, payment aggregators, prepaid instrument issuers, white label ATM operators and a handful of others. If your company is not on that list, the document does not name you.
It reaches you anyway. A company that sells a fraud engine, a KYC API, a ledger, a notification service or a hosted checkout to any of those operators is what the directions call an unregulated entity in the digital payments chain. Paragraph 5 tells the operator to make sure you adhere to the directions, and the audit section tells the operator's auditor to check that it did. The practical effect is a contract addendum, probably arriving with your next renewal.
This piece is for the founder or engineering head at such a vendor. It covers what changed on 5 November, which six clauses travel into vendor contracts, and where to push back before signing.
What changed on 5 November, and what did not
Most of the document is not new. It consolidates the Master Directions of 30 July 2024 with two older instruments: the December 2019 cyber security controls for third-party ATM switch providers and the January 2020 circular on the scope of system audits. Paragraph 58 repeals six circulars, the oldest dating from December 2009. Operators now have one rulebook of roughly 32 pages instead of seven documents.
The commencement clause has drawn the most attention. Large and medium operators must comply from the day of publication, with no transition period. That reads as abrupt until you check the 2024 text, which gave large operators until 1 April 2025 and medium operators until 1 April 2026. Both dates have passed. The new directions simply stop pretending those operators are still in transition. Small operators keep the date they already had, 1 April 2028.
So the real news for an operator is narrow: the ATM switch and system audit rules now sit inside a Master Direction issued under Section 10(2) read with Section 18 of the Payment and Settlement Systems Act, 2007, and conflicting older instructions give way to it. Early commentary from law firms describes a consolidation and none we read itemises new controls. For a vendor, the news is different. Your customer's compliance team has a fresh, single document to hold against your contract, and an annual audit that will ask to see it reflected there.
How the RBI cyber resilience directions reach a company RBI does not regulate
Three provisions do the work.
Paragraph 5 says operators shall ensure adherence by unregulated entities in their payments chain, and gives payment gateways, third-party service providers, vendors and ATM switch providers as examples. It adds two qualifiers. The adherence is subject to mutual agreement, and the operator's Board must approve a policy for it.
Paragraph 19 covers vendor risk. Operators follow RBI's August 2021 outsourcing framework, keep controls that stop an intrusion crossing from a vendor's environment into theirs, hold vendors to the rules on where payment data is stored, and obtain certified assurance from an independent auditor on the cyber resilience of any vendor involved in a critical process.
Paragraph 38 lists twenty areas the annual system audit must cover. Item (xi) is vendor and third-party risk, and it states that the service level agreement should mandatorily include a right of audit or inspection by regulators.
"Subject to mutual agreement" matters, but less than vendors hope. It confirms that RBI is not asserting jurisdiction over you and that the terms are negotiable. It does not relieve the operator, whose auditor must test vendor risk for both design and operating effectiveness. An operator that cannot get the terms has an audit observation to close within three months, and the simplest way to close it is to change vendor.
The six clauses that will arrive in your contract
| Paragraph | Operator obligation | Likely vendor clause |
|---|---|---|
| 22(d) | Report unusual incidents to RBI within 6 hours of detection; cyber incidents to CERT-In as well | Notify the operator within a fixed, short window and supply the facts its report needs |
| 38(xi) | SLA must include a right of audit or inspection by regulators | RBI may inspect your relevant systems and records, on-site or off-site |
| 19 | Certified independent assurance for vendors in critical processes | Annual third-party audit report, shared with the operator |
| 17(c) | Obtain source code of critical procured applications, or an escrow arrangement | Source code escrow with defined release triggers |
| 18(b) | Developer certificate that code is free of vulnerabilities, malware and covert channels, renewed on any change | Signed attestation per release |
| 27(e) | Preserve audit logs for at least five years | Log retention and export terms matching five years |
1. Incident notice
The operator must report unusual incidents to RBI, in the Annex 1 format, within six hours of detecting them. The text names cyber-attacks, outages of critical systems, internal fraud and settlement delays. Note that an outage qualifies. If your API is in the operator's payment path and goes down, that may be reportable whether or not anyone attacked you.
2. Regulator audit rights
This is the clause vendors resist most and operators can concede least, because the audit scope names it. Expect to negotiate its boundaries, not its existence.
3. Independent assurance
If your product sits in a critical process, the operator needs a certified auditor's view of your cyber resilience. The directions do not name a standard. A current SOC 2 Type II or ISO 27001 certificate may satisfy some operators, while others will ask for a report from a CERT-In empanelled auditor because that is who audits them.
4. Source code or escrow
For critical applications bought from a third party, the operator must obtain the source code or, failing that, put an escrow arrangement in place so the service can continue. For hosted software, source code alone does not keep a service running, so expect questions about deployment artefacts and data export too.
5. The developer certificate
Where the operator does not own the code, it must hold a certificate from the developer that the application is free of vulnerabilities, malware and covert channels, with a fresh certificate on any change to the source code.
6. Log retention
Audit logs must be kept for at least five years with controlled access. If your default retention is 90 days or a year, the operator will either ask you to extend it or ask for a continuous export it can store itself.
The six-hour clock is your customer's, so yours is shorter
The six hours run from detection by the operator. Inside that window the operator has to learn of the incident, decide it is reportable, fill in the Annex 1 format and get sign-off. Every minute you take to tell them comes out of that budget. A contract that asks you to notify within one or two hours of your own detection is not the operator being unreasonable. It is arithmetic.
Two details deserve attention before you agree to a number. First, define what starts your clock. "Detection" could mean the first alert, the moment an on-call engineer acknowledges it, or the point at which someone confirms customer impact. Those can be an hour apart at 3 a.m. Second, you probably carry a six-hour duty of your own already. CERT-In's April 2022 directions require service providers, intermediaries, data centres and body corporates to report specified cyber incidents within six hours of noticing them. One incident can therefore produce three reports: yours to CERT-In, the operator's to CERT-In, and the operator's to RBI. They should not contradict each other, which is an argument for agreeing a shared incident template with your customer now.
Two clauses that collide with how software is shipped
The developer certificate assumes software arrives in versions. Read as written, paragraph 18(b) requires a new certificate whenever the source code changes. A team that deploys twenty times a day cannot issue twenty signed letters, and an operator's auditor cannot review them. The workable answer is to define the unit in the contract: a certificate per tagged release or per quarter, backed by named evidence such as static analysis output, dependency scan results and the latest penetration test, with an obligation to re-certify on material changes. Get that definition agreed in writing, because the default reading is every change.
The wording of the certificate needs the same care. No honest engineer can state that an application is free of vulnerabilities. They can state that it was tested by specified methods on a specified date and that no known vulnerability above an agreed severity remained open. Ask counsel to draft the certificate around what you can prove.
Escrow has a similar gap. The provision was written with licensed, on-premise applications in mind, where source code plus a build environment is enough to keep going. For a multi-tenant service the operator's continuity depends on infrastructure definitions, configuration, and its own data in a usable format. A vendor that offers a tested data export and a documented exit plan alongside escrow gives the operator's Board something it can defend to an auditor.
What to do before the addendum arrives
- Find out which of your customers are authorised non-bank operators and which category each falls in. Payment aggregators are classed as large, so they are in scope today.
- Decide whether you are in a critical process for each. That single judgement drives the assurance, escrow and certificate requirements.
- Measure how long you take from first alert to customer notice on a real incident. Offer a notice window you have met in practice.
- Draft your own regulator-audit clause: reasonable notice where the regulator permits, scope limited to systems serving that operator, confidentiality for other customers' data, and cost allocation.
- Check your log retention against five years and price the storage before a customer asks.
- Prepare one evidence pack, with audit report, penetration test summary, release attestation template and exit plan, so each operator's questionnaire is a mapping exercise.
Small prepaid issuers and instant money transfer operators have until 1 April 2028, and an issuer that grows into a higher category gets a year from recategorisation. Vendors should not read those dates as their own. A small issuer usually runs on a larger operator's rails, and that larger operator is in scope now.
RBI's letter also asks operators to keep moving towards the latest security standards over time. The 2026 text is a floor that will rise, and the vendors that treat this renewal cycle as the moment to build a repeatable answer will find the next revision costs them a week instead of a quarter.
Frequently asked questions
Related reading
Optimizely Served Malware to 178 Organisations. A Content Security Policy Blocked It for Only 24.
On 16 November, Optimizely snippets delivered a card skimmer and crypto clipper to 178 organisations. Report URI checked their CSPs: 68 had none, 50 had one that allowed everything, and 24 blocked the payload.
Microsoft 365 Went Down With No Traffic Spike. Load Balancer Health Checks Need a Removal Limit.
Microsoft said load rose because servers were removed, not because traffic grew. Here is how health checks turn into a cascade, and the removal floor each major load balancer offers.
Zero Data Retention Covers a Request Path, Not a Vendor. The Fireworks Breach Reached the Paths Outside It.
The Fireworks AI breach exposed Notion AI transcripts despite a zero data retention commitment. The promise applied to inference requests. The transcripts sat in a feature that stores by design.