TL;DR

  • On November 19, 2026, JPMorganChase, Visa and Mastercard stated that a third-party services provider was breached. All three say payment data and their own systems were not affected, however the exposed customer data is marketing and contact information.
  • No services provider or vendor was named.. A group calling itself n0n claimed the data, posted files on November 16, and claimed an intrusion at Qubole, a cloud data platform. The Qubole link is unconfirmed and the leaked files are unverified.
  • Separately, Wiz, Sentra, and Mandiant documented stolen Qubole credentials being used to reach customer environments through trusted integration access, while much of it is dormant.
  • Regardless of vendor specifics or ownership of the breach, the pattern is standing third-party access which nobody monitored or governed was the root cause. In SaaS, that means OAuth grants, connected apps or AI, and integration users.
  • Two controls can apply to reduce, or nullify, the impact: limit what those integrations can reach, and alert when they start behaving differently. We cover both for SaaS applications.

What happened

Three of the largest providers and brands in consumer and enterprise payments disclosed third-party incidents within a day of each other, and each describes the exposure as limited to marketing or contact data. This is what the companies themselves said, per ABC News and Reuters (via Claims Journal).

CompanyWhat it saidData it acknowledgedPayment data and services impact
JPMorganChaseThird parties that provide services to the firm were the target of a cyberattack.Marketing data held by those vendors, which may include some customer names and email addresses.Payment services unaffected. No disruption to its services.
VisaThe claims are associated with an incident at a third-party service provider in Asia Pacific.Not specified.Visa’s systems were not involved. No payment data such as card numbers was affected.
MastercardAn incident at a third-party provider exposed a limited number of email addresses.Email addresses for a small subset of cardholders in one benefit program.No impact to the payment network. Exposure contained.

The public timeline, in order: 

  • November 11, 2026, n0n, an extortion group that emerged in September 2026, claimed an intrusion at Qubole. 
  • November 16, 2026, it posted files it said came from suppliers serving the three companies. 
  • November 19, all three companies acknowledged third-party incidents.

Reuters reviewed the posted data and could not verify it. Its reporting describes more than 16,000 customer email addresses and about 124 bank employee addresses in the JPMorgan dataset, with some phone numbers and postal codes. Treat those figures as claims until a company confirms them.

JPMorgan also stated the breach has no connection to the service disruptions which some customers saw earlier in the week, and JPMorgan attributed that separate occurrence to a software bug.

What has been reported about how it happened

The impacted companies have not said how their vendors were breached, and the reporting does not establish it. Here is what the suspected actor claims, and what researchers documented separately:

Claimed by n0n 

The group says it breached Qubole, a cloud data analytics platform, and holds data from its customers, including a claimed nearly one petabyte that is unverified. BleepingComputer reports Qubole as suspected and not responding. Malwarebytes notes that none of the three companies confirmed Qubole as the vendor involved. Cybernews researchers think the leaked files share formatting that points to one common provider, which is an inference from file structure.

Documented by Wiz 

In a separate report, Wiz, working with Sentra and Mandiant, describes an attacker using credentials from an AWS account associated with Qubole between October 14 and 26 to assume roles that customers had created to trust Qubole. Every request carried the correct External ID and succeeded, and most of the roles were not otherwise in use. The actor then listed resources, pulled secrets from Lambda environment variables, Secrets Manager and Parameter Store, tested what it found against AWS, Entra ID and Okta, and in some cases downloaded entire S3 buckets. Wiz does not say how Qubole’s credentials were obtained and does not mention JPMorgan, Visa or Mastercard.

What we still do not know:

  • Whether the JPMorgan, Visa and Mastercard data came from Qubole at all.
  • Whether one vendor explains all three disclosures.
  • Where the exposed marketing data lived, whether in cloud storage, a SaaS marketing platform, or elsewhere.
  • When the bank vendors were breached and how many customers are affected.

What could prevent or detect it

Stolen vendor credentials used against a legitimate trust relationship look like the vendor, so a login alert or a perimeter tool is unlikely to flag them. The controls that are applied implement limitations to the trust, what it can access, and watch how it is used. They follow the identify, protect, detect, respond cycle.

  • Identify. Inventory third-party access in every SaaS application, including marketing and customer-engagement platforms: connected apps, OAuth token grants, integration users and service accounts. Mark each as active, unused, or unknown. Dormant access was central to Wiz’ findings  about Qubole customers.
    • A count of grants is not a priority list. A dormant grant held by an admin with full-access scope and no IP restriction is a different risk from a dormant read-only one. Score each one on what it can reach and how exposed it is, and start at the top.
  • Protect. Revoke stale grants and reduce scope to least privilege. Make sure no integration runs as a privileged user without a reason. Restrict the networks an integration can connect to, and continue audit logging on and retain the log data long enough to review and audit if needed. We suggest one or two years.
  • Detect. Alert on integration behavior, not on attacker-controlled strings: a first-seen network or user agent, access from VPN or hosting infrastructure for an integration that normally does not use it, and bulk read behavior. Wiz saw Mullvad VPN and OVH-hosted addresses. VPN exits are shared, so this needs a baseline per integration.
  • Respond. Know how to revoke a grant, disable an integration and rotate its secrets within hours, and revoke active sessions and tokens as well as rotating credentials.

Also ask third-party vendors where they store your marketing and customer contact data, and how and how soon they would notify you about a potential impacting breach. For AWS environments, the Wiz write-up covers role-trust hardening and CloudTrail hunting.

How AppOmni can help

We cover both halves of the above list for SaaS and AI applications. Posture rules show whether an integration is set up in a way an attacker could abuse. Threat detections alert when something uses that access abnormally. A posture finding says a door is open, and a detection says someone walked through it.

With AppOmni’s risk-based posture experience each violation receives its own 0 to 100 risk score customized from your environment, with each factor explained, so two stale OAuth grants no longer look the same. Stale or Unused OAuth Token Grant weighs how long a grant has been idle, whether its token was ever used, and whether the user who granted it is still active. This methodology would detect the dormant, trusted access Wiz found at Qubole customers. Teams start with the few grants that matter. A grant accepted as low risk stays visible, and if its score climbs past a threshold you set, it reopens automatically. AppOmni correlates posture issues, activity signals, and threat alerts to show teams which risky grants to fix first.

The examples below are Salesforce rules and detections:

RiskPosture rule (close the door)Detection (see someone use it)
Stale integration accessStale or Unused OAuth Token GrantOAuth App Login from VPN and Anomalous Network Location; High Volume API Query from VPN Address
Weak network restrictionsOverly Broad Trusted IP Range; Overly Broad Profile Login IP RangeOAuth App Login from VPN and Anomalous Network Location
Over-privileged integrationExternal Client App Client Credentials Flow Runs as Privileged User; Bot or Agent Runs as Over-Privileged IdentityRisky Scopes Assigned to App; Mass API Queries from First-Seen IP Address and User-Agent
Bulk data pulls through an integrationDetection-ledMass API Queries from First-Seen IP Address and User-Agent; Bulk API Query from Anomalous Network and Network Location
Not enough logging to look backSecurity Monitoring and Audit Logging Controls DisabledTransaction Security Policy Modified

What’s coming next from AppOmni:

These upcoming rules extend the same coverage and are not yet generally available. 

  • Connected App With Weak OAuth Policy (Salesforce) flags apps that bypass org IP restrictions, let any user self-authorize, grant full-access scope or skip PKCE. 
  • Application Credential With Excessive Lifetime and Application With High-Privilege Microsoft Graph Application Permission (Microsoft 365) cover long-lived Entra ID app secrets, the kind the attacker harvested in the Wiz report, and how far one can reach. 
  • Valid OAuth Token for Dormant or Deactivated User (ServiceNow) applies the dormant-grant check beyond Salesforce.

To see how AppOmni can help your enterprise, please request a demo here: https://appomni.com/demo-request/.

If you would like to talk directly about how AppOmni can secure your enterprise, please contact sales@appomni.com.

Sources