Feature Incident Response
The Outage Ended at 7 p.m. The Phishing Wave Runs for Two Weeks.
Post-outage phishing: the callback-to-a-known-number control, a 14-day fraud package with owners, and what your own outage notice must contain.
Table of Contents
TL;DR
- The Chase software bug was resolved by about 7 p.m. on November 16. The impersonation volume peak comes after that, not during it — Halcyon’s Cynthia Kaiser expects criminals to work the outage pretext “for the next few months (at least!).”
- Harris County law enforcement started warning residents about phishing texts, calls, and emails exploiting the outage the same week (FOX 26 Houston). The control that survives caller ID spoofing is a callback to a number the customer already had.
- Run the incident for fourteen days past restoration, on a schedule with owners. Restoration is a systems milestone, not an incident close.
At 2:05 p.m. on Monday, a customer re-taps a stalled Zelle payment, gets an error, and then gets a text: “Chase: your transfer failed. Verify your account to release funds.” The link is short. The logo is right. The timing is perfect.
The outage handed fraud teams something they usually have to manufacture: a reason to believe. Every customer on the rail had just been given proof that their bank was broken, so the text was not an interruption — it was a status update.
By Wednesday, Harris County Precinct 4’s constable’s office was telling residents to stop clicking. Captain Jonathan Zitzmann’s rule, in his own words: “don’t respond to it, don’t acknowledge it, certainly don’t click on any links that may be included in a text message or an email” (FOX 26 Houston).
Then he gave the version that actually works. Turn the card over. Call the number on the back. Verify through a channel the attacker does not control.
The pretext writes itself
Craig Timm, senior director of anti-money laundering at ACAMS, told CBS that consumers should “assume they will be targeted,” and named the tell: pressure “to take urgent action to protect your funds or restore account access” (CBS News).
Cynthia Kaiser, SVP at the Halcyon Ransomware Research Center and a former FBI cyber deputy director, put the durability of it bluntly to Cyber Magazine: “Make no mistake, criminals and adversaries are waiting to take advantage of the situation.” Her forecast is a long tail, not a spike — “for the next few months (at least!) most organisations will be fixing security flaws they did not know existed. We’ll see more outages.” (Cyber Magazine).
Zitzmann also noted he had personally taken a fraudulent call from someone impersonating him. Spoofing is cheap, and the busy-signal window of a live outage means fewer customers can check.
The control that survives a spoofed number
Most banks have a callback rule. Most of them are written to sound reasonable and fail at the desk, because they say “verify the caller” without saying against what.
Write it as three lines with a stop:
- Identity is established from a number already on file — statement, card back, app profile, or a prior documented call. Never from the inbound call.
- The customer is told to hang up and call the printed number. If they cannot, staff call back on the number on file while the customer holds.
- Stop line: no one-time code, full card number, or password is accepted on the inbound call until step 1 or 2 completes. If the caller pushes, the answer stays no.
The stop line is what makes it enforceable. Frontline staff under pressure will accept a code from a persuasive caller to clear the queue; naming the code as unacceptable until verification gives them a line to stand on.
FFIEC expects the surrounding architecture. Section III.D requires defined protocols to declare and respond to incidents, explicit authority delegation for containment and restoration actions, external communication planning, and documentation of decisions, with a program that is periodically tested.
Fourteen days, with owners
Restoration is when the operating tempo should go up, not down. The bundle of outage, pretext, and impersonation is the standard sequence — it is why the incident does not close on the app going green. A workable post-restoration package, starter-shaped and to be sized against your own volumes:
| Day | Control | Owner | Artifact |
|---|---|---|---|
| 0 | Freeze duplicate resubmissions; open exception queue | Payment Ops | Exception log with transaction IDs |
| 0–1 | Publish one verified-contact notice on owned channels | Comms + Fraud | Notice ID, canonical URL |
| 1–3 | Callback script refresh and 15-minute desk huddle brief | Contact Center Ops | Script version, huddle attendance |
| 1–14 | Imposter-call tagging on every inbound interaction | Fraud Ops | Tag code, daily count |
| 2–14 | ATO rule uplift: new-device plus rail-tagged failure = step-up (playbook) | Fraud Strategy | Rule change ticket, false-positive rate |
| 2–14 | Review disputed and reversed transactions tagged to the window | Disputes | Case sample, outcome codes |
| 3–10 | Outbound sweep to customers with a failed rail transaction | Customer Care | Contact list, disposition |
| 7 | First lessons-learned entry, including queue abandonment | Incident Lead | Memo section |
| 14 | Close-out memo with residual exposure and rule rollback decision | CISO or delegate | Signed memo |
The line that decides whether this works is the outbound sweep. An institution that waits for customers to call absorbs the losses; one that calls the customers with a failed transaction can catch a takeover in progress, because the attacker is calling the same people.
Your own notice is an authentication artifact
Customers cannot tell your message from a fake one unless yours has a fingerprint.
Build the fingerprint in deliberately. Every outage notice names the specific rail or service, states the current status, gives one instruction, names the one thing you will never ask for on an inbound contact, and carries a notice ID posted at the same URL on your own domain. That is the reference a customer or a branch can check a suspicious message against.
Say the negative out loud: we will not ask for a one-time code, a full card number, or a password by text, and any call can be verified by calling the number on your card.
Chase’s customer-facing line during the event — “some services aren’t available. We’re working on them. Please try back later” — did the first two jobs. The third is the one fraud teams usually forget to add.
Where this breaks
Queue abandonment. During the outage, real calls back up, hold times climb, and agents get faster with each one. That is when a code gets accepted. Watch abandonment more than wait time; abandonment is the proxy for “agent improvised.”
The helpful agent. Someone trying to be useful will offer to “verify it for you right now.” The callback control should be phrased so the helpful path and the compliant path are the same path.
Duplicate resubmission. A customer who re-sends a stalled transfer creates a genuine double payment, then gets two debits and a fraud case that is not fraud. The freeze order matters more than the freeze itself — publish it before 3 p.m., not at close.
Four metrics that move early
Track daily for fourteen days, alongside a pre-event baseline from the prior thirty:
- Imposter share of inbound contacts — tagged contacts divided by total. Starter amber around double baseline, red at triple.
- Callback abandonment — share of customers who hang up before a call-back completes. Amber above 10 percent.
- Step-up challenge rate on new-device transactions — a sharp fall means the rule is not firing.
- Dispute rate on rail-tagged transactions — compare to the same weekday for four weeks.
Calibrate all four against your own history over three to six months before publishing them as thresholds. Anti-gaming check: sample twenty imposter-tagged contacts weekly and confirm the tag matches the transcript, and re-open any case where an agent accepted a code before verification.
So what
The FFIEC booklet’s restoration requirements end on a sentence most teams skim: monitor for similar or further incidents. The phishing tail is that sentence. It also asks for a Suspicious Activity Report decision where applicable — impersonation-driven takeovers usually land there.
Practical close: put a calendar item on November 20 and another on December 1, then two more. The outage is over. The incident has a fortnight left.
The Incident Response & Breach Notification Kit includes the callback script, the post-restoration monitoring schedule, and the close-out memo.
FAQ
Is this different from the 36-hour notification question? Adjacent, not the same. Notification classification is a decision about the disruption; this is the customer-facing tail. A bank can correctly conclude the outage was a notification incident and still under-run the fourteen days after — the examiner view after a notice goes out assumes the tail exists.
Who owns the close-out memo? The CISO or a named delegate — someone senior enough to also decide when to roll the step-up rules back. Fraud Strategy proposes the rollback; the incident owner signs it.
What if the volume never spikes? Keep the schedule for fourteen days anyway and record the null result. A documented quiet fortnight is a defensible answer to an examiner asking what monitoring followed restoration.
◆ Need the working template?
Start with the source guide.
These answer-first guides summarize the required fields, evidence, and implementation steps behind the templates practitioners search for.
◆ Related template
Incident Response & Breach Notification Kit
Run an incident and every notice clock it starts: bank partner, the bank regulators' 36-hour rule, NYDFS, FTC, SEC and breach laws in 54 states and territories. Workbook, guide, four playbooks, tabletop kit and Word plan templates.
◆ Immaterial Findings · Weekly
Sharp risk & compliance insights. No fluff.
◆ FAQ
Frequently asked questions.
Why does fraud rise after a public outage?
What is the callback control, in operational terms?
Does a five-hour outage that was not a cyberattack still need incident response?
What should an outage notice include so customers can spot a fake one?
Author
Rebecca Leung
Rebecca Leung has 8+ years of risk and compliance experience across first and second line roles at commercial banks, asset managers, and fintechs. Former management consultant advising financial institutions on risk strategy. Founder of RiskTemplates.
◆ Related framework
Incident Response & Breach Notification Kit
Run an incident and every notice clock it starts: bank partner, the bank regulators' 36-hour rule, NYDFS, FTC, SEC and breach laws in 54 states and territories. Workbook, guide, four playbooks, tabletop kit and Word plan templates.
◆ Keep reading
Related posts.
Incident Response
The Memo You Sign Before the 36-Hour Clock Starts
The 36-hour clock runs from your determination, not detection. Here is the notification incident determination memo, who may sign it, and how to record a holding answer.
Nov 16, 2026
Incident Response
NCUA's 2026 Cybersecurity Report: Most Credit Union Incidents Involve Vendors the Regulator Can't Examine
The NCUA's 2026 Cybersecurity and Credit Union System Resilience Report documents 588 reported incidents, including third-party provider incidents at vendors NCUA has no authority to directly examine. NCUA's earlier data put nearly 7 in 10 incidents at a vendor, and one 2023 ransomware attack on a service provider knocked out 60 credit unions. Here's what this structural gap means for your incident response program.
Oct 7, 2026
Incident Response
Your Incident Response Plan Has the Wrong Trigger. Here's What Starts the SEC's 4-Business-Day Clock.
Two years of SEC Item 1.05 cybersecurity disclosures reveal one consistent mistake: IR plans built around incident discovery, not materiality determination. Here's what the trigger actually is — and how to fix your escalation process before it matters.
Oct 6, 2026