advertisement
The Crisis Nobody Had To Declare
The most revealing thing Equity Bank did last week was publish a phone number.
With its email down and customer enquiries piling up somewhere inside Microsoft’s infrastructure, the bank told customers to reach it by telephone or WhatsApp instead. Other services, it said, were running normally. It was a sensible, quick, practical response to a problem the bank had not caused and could not fix.
It was also an admission, made in public and without anyone quite registering it, that a regulated financial institution’s fallback channel for talking to its customers during a service disruption is a consumer messaging app it does not own, cannot audit and has no contract with.
advertisement
The same scene played out across the continent in different accents.
In Cape Town, the city administration told residents that staff might be unable to send or receive email, that Teams connections were intermittent, and that “at this stage, there is no confirmed timeline for the full restoration of services.” The South African Revenue Service warned that the outage “may affect access to Outlook, Teams and other Microsoft 365 applications” and said its technical teams were engaging directly with Microsoft. The University of Cape Town passed on the same warning to students and staff.
The failure began around 8 pm in Johannesburg on Thursday 12 November, which is 9 pm in Nairobi. Late enough to miss the working day in both cities and early enough to ruin the Friday morning that followed. Reports spiked after 5 am as people logged on and found nothing there. It ran well past 18 hours, and Microsoft did not declare it resolved until 14 November. The company has since attributed it to an issue introduced during an earlier network capacity upgrade. There was no attacker.
advertisement
Which brings up the question worth asking now that the inboxes have drained. At what point does an outage stop being an IT problem and become a business crisis?
There are three possible answers.
The Contract Will Not Tell You
Start with the document everyone assumes covers this. Enterprise cloud agreements carry a service level commitment, and when the provider misses it, the remedy is a service credit. A proportion of the monthly fee for the affected service, which the customer has to notice, calculate and claim within a short window.
advertisement
A credit is a refund. It is not compensation. It bears no relationship to invoices that were not sent, approvals that were not given, theatre lists that slipped or a trading desk that spent the morning on the phone. The provider’s maximum exposure is a fraction of what the customer was already paying it. The customer’s exposure is the whole business day.
That asymmetry is not a scandal; it is how the market is priced, and it is part of why the service is affordable. But it does mean the contract is silent on the only question that matters to a board. It defines when the provider owes you money. It says nothing about when you are in trouble.
The scale of the gap has been measured, if loosely. Splunk’s research with Oxford Economics, published this year, put the cost of downtime across the Global 2000 at $600 billion a year, roughly $300 million per organisation, or about $15,000 a minute. Figures from a vendor whose business is selling observability, and worth reading with that in mind. Even discounted heavily, no service credit regime operates anywhere near that order of magnitude.
Lagos Drew The Sharpest Line
The second answer is the regulator’s, and here the continent does not speak with one voice. It does not even agree on what kind of thing an outage is.
Nigeria has the clearest trigger, which will surprise people who have not read it. The Central Bank of Nigeria’s Risk-Based Cybersecurity Framework for Deposit Money Banks and Payment Service Banks defines a cyber-incident by what it does, not by who caused it. The list in Appendix I includes an “unplanned outage in Core Banking Application” and “any glitch that results in financial losses.” Appendix VII, the reporting template itself, lists “Unplanned Downtime” and “System Failure” as incident types. Section 5.0(v) requires institutions to “report all cyber incidents… not later than 24 hours after such incident is detected” to the Director of Banking Supervision.
There is no requirement that anybody meant it. A power failure, a botched upgrade, a provider’s bad night; if it takes the core banking application down, the clock starts.
Nigeria also went further than most of its peers on the structural question. Section 3.7 directs institutions to “avoid creating single points of failure in the industry” and to plan for mitigating that risk. That is a concentration-risk clause, written into a banking supervision document, which is precisely the exposure last Thursday demonstrated. The country has had its own rehearsal: in October 2024, a power failure at a MainOne data centre took a number of Nigerian banks offline, with no attacker anywhere in the story.
South Africa has arrived at the same place by a different route. Joint Standard 2 of 2024, issued under the Financial Sector Regulation Act, defines a cyber incident as an event that jeopardises the cybersecurity of an IT system or violates security policies — “whether resulting from malicious activity or not.” Those last six words do a great deal of work. Cybersecurity, as the Standard uses it, includes preserving availability. And the reporting duty in paragraph 9.1 attaches to a material incident, which the Standard defines purely by effect: “a disruption of a business activity, process or function” that “has, or is likely to have, a severe and widespread impact” on operations, customers, or the financial system.
Cause is irrelevant. Impact is the test. The weakness is at the other end: the Standard sets no notification deadline, leaving the form, manner and period to be determined by the authorities. South African institutions know what counts. They are still waiting to be told how fast.
Ghana has been moving in the same direction. The Bank of Ghana launched a revised Cyber and Information Security Directive this year, replacing the 2018 original, requiring every regulated institution to have at least one board member with verifiable cyber risk expertise and restricting where sensitive financial data may be hosted. Governor Johnson Pandit Asiama framed the purpose in terms that fit last week exactly, safeguarding “the confidentiality, the integrity and the availability of the data.” Availability is in the sentence. Whether the directive’s reporting triggers reach a non-malicious supplier failure is not yet clear from what the central bank has published.
In Nairobi, The Trigger Is Still Malice
Against that, Kenya looks like the outlier, and the gap is not a technicality.
The Central Bank’s Guidance Note on Cybersecurity for the Banking Sector requires an institution to “notify the Central Bank of Kenya within 24 hours of any Cybersecurity incident(s).” Clear duty, firm clock, and Kenyan banks take it seriously.
The difficulty is the definition. A cybersecurity incident, under the same document, is “any malicious act or suspicious event” affecting critical cyber assets. Microsoft has said repeatedly that there was no malicious activity. Charles Lamanna, the company’s Executive Vice-President for Copilot, agents and platform, put it plainly on Thursday night that “there is no indication this is the result of malicious or cyber activity.”
So, by the Guidance Note’s own terms, a failure that took email away from several of Kenya’s largest banks, pushed customer contact onto WhatsApp, and disrupted hospitals and investment firms was not a reportable incident. The same event, in a Lagos bank, would have started a 24-hour clock the moment it was detected.
The Guidance Note is not blind to the underlying risk. Its definition of cyber risk covers failures of IT systems that cause “disruption of services,” and section 3.3 asks banks to monitor third parties and include availability provisions in their service level agreements. But that is a supervisory expectation, not a notification duty. Nothing in it starts a clock.
The data protection route closes too. The Data Protection Act’s notification obligation attaches to a breach of personal data, not to an inability to reach it. Nothing was exfiltrated. Nothing was exposed. The information was simply unavailable to the people entitled to it, which the law does not treat as an event anybody needs to be told about.
The document is eight years old, written when the risk that kept supervisors awake was an intruder rather than a dependency. That is not a criticism of its drafters. It is an observation about how quickly the problem changed.
The Threshold That Actually Binds
Which leaves the third answer, the operational one, and it is the only test that held up on Friday.
An outage becomes a crisis when the system that’s down is the one you’d use to report it, like email going down when email is how you warn everyone.
That is why Equity’s notice was the right instinct and the uncomfortable detail at the same time. The bank had somewhere to go. Cape Town had somewhere to go. Many organisations discovered on Thursday night that their incident communication plan, the agreed way of telling staff, customers and regulators what was happening, existed largely in Teams, or in Outlook, or in a distribution list that existed only inside the platform that was down. A plan that depends on the failed system is not a plan.
Walter Pasquarelli, an affiliate researcher at the University of Cambridge, reached the same conclusion while the incident was still running. “This is first and foremost a communication risk,” he said.
The dependency chains ran deeper than most organisations map. Reports came from 74 countries, according to Downdetector. Government systems in Singapore and Japan were affected, as were communications at the Supreme Court of the Philippines and, in Australia, the federal police. Third-party tools built on Microsoft infrastructure were caught in it, including services from OpenAI and Anthropic. Which is to say that organisations which had carefully avoided depending on Microsoft directly found they depended on it anyway.
Britain’s response to this class of risk has been to pull the supplier inside the fence. In July, HM Treasury designated four companies as critical third parties to the UK financial sector: Amazon Web Services EMEA, Google Cloud EMEA, Oracle Corporation UK and Microsoft Ireland Operations Limited, so that the regulators’ rules now apply to them directly for the systemic services they supply to finance. The entity whose infrastructure failed last Thursday is, in London, a supervised one. In every African market it serves, it is a vendor with a service credit schedule.
What To Watch
Three things will show whether last week changed anything.
Whether any Nigerian institution filed a 24-hour report, given that the framework appears to require one and the incident is exactly the kind it describes. Whether the South African authorities finally attach a deadline to Joint Standard 2, which would turn a well-drafted definition into an operative duty. And whether the Central Bank of Kenya revisits a definition written around intent, when the risk that stopped work last week had none.
Underneath all three sits the question any board on the continent can ask its executives this week, and it is narrow enough to answer in a meeting. If our primary platform is unavailable for 18 hours, what is the agreed channel for telling customers, who owns it, and is it something we have actually contracted for?
If the honest answer is a consumer messaging app and a switchboard, that is worth knowing before the next one. And there will be a next one, because the fix for this outage was applied inside a company most African organisations will never meet, under terms they did not negotiate, on a timetable they could not influence.