What the companies have confirmed
J.P. Morgan Chase said third parties that provide services to the firm were targeted in a cyberattack. It believes the attackers accessed certain marketing data held by those vendors, which may include some customer names and email addresses. The bank said its services weren't disrupted and that the incident is unrelated to a service disruption earlier this week. Chase did not name the vendors in the statement quoted by BleepingComputer.
Visa and Mastercard each described a separate third-party incident:
- Visa: Visa told BleepingComputer the claims relate to a third-party service provider in Asia Pacific. It said its systems weren't involved or affected and that no payments data such as card numbers was exposed. Visa told Newsweek the same thing about the Asia-Pacific provider and its own systems.
- Mastercard: Mastercard said a third-party provider incident exposed a limited number of email addresses. These belonged to a small subset of cardholders enrolled in a specific benefit program. It said the payment network and card use are unaffected and that the exposure has been contained.
None of the three statements describes a breach of the banks' or networks' own core systems. Visa and Mastercard have not confirmed any impact on customers.
Who is claiming it, and what is unproven
On November 11, 2026, the n0n group claimed responsibility for an attack on Qubole, which BleepingComputer says is suspected to be the third-party provider. n0n published data it claims belongs to Chase, Visa and Mastercard on its extortion site on November 16.
BleepingComputer describes Qubole as a cloud data analytics company owned by Idera. The group allegedly stole 53.9 GB across 30,743 files directly from Qubole. It also claims to hold nearly a petabyte of data from Qubole customer networks. These are the attacker's own figures, and nobody has verified them. The same applies to the claim that Qubole is the common link between the three financial firms.
Check Point's Cyberint unit takes a similar position. It assesses with moderate confidence that the claims trace back to one or more shared data vendors rather than direct intrusions into the banks or card networks. It lists Qubole as a possible source but says there is currently no public evidence to confirm that.
Reports on the scale of the leak differ and are not confirmed counts. Newsweek says n0n published around 16,000 customer email addresses and phone numbers. It also reports that Cybernews researchers found Visa-related data on more than 6,000 premium cardholders. The companies themselves haven't given figures. Chase's 78 million digital customers and $4.1 trillion in assets are background only. They say nothing about how many people were affected.
The technical story: cross-account AWS trust
Wiz, working with Sentra and Mandiant, published technical findings on the Qubole-related AWS activity. Wiz says it observed a threat actor using IAM credentials from an AWS account associated with Qubole between October 14 and 26, 2026. The actor used those credentials to assume IAM roles that customers had created to trust Qubole's account. Wiz did not independently tie this activity to the datasets n0n published. It also says it can't see the actor's motivation or what the actor intends to do with the data.
How the trust model works
SaaS platforms that work inside a customer's AWS account usually ask the customer to create a role. The role's trust policy lets the vendor's AWS account assume it, typically gated by an External ID. The vendor then receives temporary credentials scoped to that role.
In this campaign, every role assumption Wiz attributed to the actor included the correct External ID, and every one succeeded. Wiz notes that most of the roles involved weren't otherwise in use. In most environments Wiz observed, the actor's requests were the only AssumeRole activity from the vendor account in October. The trust policies were still active, so the requests worked.
What the actor did once inside
- Discovery: Session names followed a
quiam-prefix, such asquiam-enumandquiam-scope-full. The actor enumerated storage, compute, databases, identity and keys. - Credential harvesting: The actor read Lambda function configurations, which can expose environment variables. It also pulled Secrets Manager values and SSM Parameter Store parameters.
- Credential testing: The actor used TruffleHog to validate discovered credentials. These included AWS access keys, Microsoft Entra ID service-principal client secrets and Okta client credentials.
- Data theft: The pattern was to assume a role, list buckets, list and inspect objects, then download them. Wiz describes the actor as opportunistic. In some cases it downloaded entire buckets in one session.
The Entra ID detail matters to Microsoft-centric shops. A secret sitting in a Lambda variable or a parameter store entry can unlock identity-provider access well beyond the AWS account it was found in.
Indicators Wiz published
- AWS account: 805246085872, the source of the role assumptions. It also carries legitimate Qubole platform traffic.
- IAM principals: AIDA3W7DIFLYOLTK63SFS, which Wiz saw only in actor activity, and AIDAJYRRH5Y3A7AAJM2HW, which also carries legitimate traffic.
- Network addresses: Mullvad VPN exits and an OVH-hosted server at 57.129.54[.]120. Wiz warns that shared VPN addresses alone are not conclusive.
- Legitimate traffic: Wiz says all the legitimate platform traffic it saw came from AWS infrastructure. A role assumption from the vendor account via a non-AWS address is therefore a strong signal.
What defenders should do
If your organization integrated with Qubole, Wiz's guidance covers both hunting and cleanup.
- Hunt: Review CloudTrail from October 1, 2026 onward in every account that holds an integration role. Look for AssumeRole events from the account and principals above that come from non-AWS source addresses.
- Check for credential testing: Look for GetCallerIdentity calls with a TruffleHog user agent paired with the listed IPs. Check your identity provider's sign-in logs for the same user agent.
- Trace each session: For every suspicious role session, review what followed, especially Secrets Manager, SSM, Lambda and S3 activity.
- Rotate and revoke: If a role was abused, rotate every secret, parameter and Lambda environment variable it could read. Revoke active sessions for the role as well, because rotating a credential doesn't end temporary sessions already issued.
Wiz's longer-term advice applies to any SaaS integration:
- Inventory every role that trusts an external account.
- Delete trust for integrations you no longer use.
- Where the vendor supports it, narrow the trusted principal and add
aws:SourceIporaws:SourceVpceconditions. Wiz says a source IP condition alone would have blocked every role assumption it observed. - Treat the External ID as non-secret. It appears in plaintext in CloudTrail and is often stored by the vendor and in infrastructure-as-code.
- Apply least privilege. Lambda configuration, Secrets Manager, SSM and broad S3 read are often granted for convenience.
- Enable an organization-wide trail, plus S3 data events or server access logs for sensitive buckets. CloudTrail management events record the role assumption and bucket listing but not the object reads that make up exfiltration. Wiz says missing logs left responders unable to see which objects were read in several affected environments.
Loose ends
- A Windows-flavored ISO: Wiz found a bootable Linux image named
sso_agent.isoon the actor's OVH server. It loads the ntfs3 driver, which suggests it targets Windows machines. Wiz says it has found no deployments and can't confirm whether or how it was used. Don't read it as evidence of a Windows compromise tied to this leak. - A separate campaign: Wiz also saw activity from October 8 to 11 using customer-owned AWS access keys it believes were created for Qubole integrations. It could not link that activity to the October 14–26 role-assumption campaign.
- Qubole's response: BleepingComputer said it had contacted Qubole and not yet received a reply.
What this means for affected customers
Names, email addresses and benefit-program membership aren't card numbers. They are still enough to make a phishing email look convincing. That is a risk inference from the data types, not evidence that phishing has happened. Treat unexpected messages that cite a card benefit or bank relationship with suspicion. Go to the company's official app or site rather than clicking a link. Neither the company statements nor BleepingComputer's report says anyone needs to replace a card.
The wider lesson is that the biggest exposure here wasn't a bank firewall. It was a cross-account trust relationship that was set up once and then left open.
References
- Vendor data breach exposes Chase, Visa, Mastercard customer info BleepingComputer · 2026-11-20T10:58:06-05:00
- Hackers Leak Data on Thousands of JPMorgan Chase’s Richest Customers - Newsweek newsweek.com
- Threat Actor Investigation n0n cyberint.com