On 16 November 2026, the n0n ransomware group, a data-theft extortion crew, listed Visa, Mastercard and JPMorgan Chase on its Tor-based leak site and started a 48-hour countdown. It claimed to hold cardmember marketing profiles, contact records, premium travel and dining bookings, and more than a million marketing-campaign interactions across several regions. Within hours the story had the shape of a standard ransomware event: a criminal crew, three household-name financial institutions, a ticking clock and the threat of a data dump.
The public record as of the morning of 18 November is messier. Visa confirmed an incident at a third-party provider. Chase blamed a same-day service disruption on a software bug and denied any cyber cause. Mastercard said nothing. No payment-card numbers have been confirmed exposed anywhere. The group itself has a documented technical profile, and I think it deserves a close look on its own terms. The listing does not prove a breach, but the tools and access paths n0n has been seen using are the ones defenders need to close however these allegations resolve.
Leak-site listing vs verified breach
A listing is an assertion. Groups often time one for maximum pressure and minimum verification, so it stays unverified until victim confirmation, forensic evidence or authenticated sample data backs it up. That is the central operational fact of this case.
In the Visa, Mastercard and Chase listings, the described datasets are marketing and cardmember-programme records. They are not card-authorisation, settlement or core-ledger systems. That does not make them harmless. Cardmember emails, programme membership labels, postal codes and campaign interaction histories are exactly what makes a targeted phishing campaign convincing. Even so, a marketing dataset sits a long way from a payment-card dump, and conflating the two misreads both the risk and the actor.

Visa: a third-party provider incident in Asia-Pacific
Visa’s public position, reported by Cybernews, is that the incident under investigation originated with a third-party service provider in the Asia-Pacific region. Visa says its own systems “were not involved and were not impacted,” and that there is “no indication that payment data such as card numbers has been affected.” It has not named the provider.
The n0n listing describes Asia-Pacific cardholder-programme records: names, personal email addresses, mobile numbers, hotel and restaurant bookings for premium cardholders, and records of more than 1.5 million marketing-campaign interactions. That fits what a programme-management or marketing-analytics vendor would hold on a payment network’s behalf. It is a plausible data-class match, which is a lower evidentiary bar than proof.
The more useful signal is the trend. Visa’s own reporting, released on 17 November 2026, said incidents involving third-party agents rose 62% in the first half of 2026 compared with the previous six months, and that AI-enabled fraud and third-party attacks were scaling worldwide. The n0n claim sits inside that documented pattern of supply-chain and agent-mediated exposure, even though nobody has independently validated it.
JPMorgan Chase: a same-day outage and a leak-site listing
Chase’s situation is legally and technically separate. On the day the listing appeared, the bank had a service outage. Downdetector logged more than 2,000 reports by 11:04 EST and more than 4,500 by the afternoon. Users also reported problems at Bank of America, U.S. Bank, Wells Fargo, Truist and Zelle. Chase told Reuters the disruption was a “temporary issue related to a software bug in on-premise hardware systems” and that there was “no evidence it is cyber-related.” Other banks may have been affected only because they were parties to transactions with Chase.
The Chase listing appeared roughly two hours and forty-four minutes before Downdetector began recording reports. Timing like that invites causal inference and proves nothing, since a criminal group can list a company on the same day as an unrelated outage. Chase’s statement denies a cyber cause for the outage. It does not deny that n0n holds Chase data, and the bank has not said whether the outage had any connection to the group’s claims. Those are separate propositions, and reporting should keep them apart.
What n0n says it holds on Chase
Reconstructed from the leak-site entry, the Chase listing claims three datasets:
📬 Stay Ahead of Cyber Threats
Get the latest cybersecurity news, critical vulnerabilities, threat intelligence, tutorials, and exclusive giveaways delivered straight to your inbox. No spam. Unsubscribe anytime.
- Engagement records for “a premium Reserve program of a global bank,” including hundreds of thousands of campaign interactions and tens of thousands of cardmember email addresses.
- The member directory behind those records, with programme labels and postal codes.
- Campaign emails whose subjects and sender addresses carry the bank’s own branding.
All of it is marketing and programme data. There is no claim of access to card authorisation, settlement or the core banking ledger. On its own, this data supports targeted social engineering far more directly than it supports card fraud. With cardmember emails, programme labels and postal codes, an attacker can write a message that looks like it came from the bank’s own programme, attempt credential stuffing against the bank or against other services where customers reused passwords, and combine the records with other breach data to make a fraud call or password-reset attempt more convincing.
Card numbers, expiry dates and CVVs are absent from the claimed data, which is a material limit. Visa’s statement that no payment data was affected covers Visa’s own systems and the third-party incident it described. It says nothing about every other dataset n0n may hold.
Mastercard: no confirmation so far
Mastercard, named in the same listing wave, has confirmed nothing. Its entry appears to describe programmes in the Middle East and Africa, a different region from the Asia-Pacific focus of the Visa listing. That weakens the simplest hypothesis, that all three listings come from one shared vendor exposure, without disproving it. The listings could reflect several vendors, data aggregated from different programme managers, or a broad claim built from whatever the group could reach. The public record does not settle it, and any analysis that treats the three listings as one coherent breach assumes more than the evidence supports.
Who is the n0n ransomware group?
n0n emerged in September 2026. By mid-October, SOCRadar had counted 93 listed victims across 27 countries, concentrated in the United States and South Korea. By early November, Check Point counted more than 170 victims and placed n0n among the top three published ransomware actors, with 8% of published attacks in October. ZeroFox observed at least 105 attacks since the group’s first victim, an average of roughly two a day, and assessed it as an immature, low-to-medium sophistication collective with a roughly even chance of becoming more dangerous over the next year.
The tempo is high and victim selection is opportunistic. The group is financially motivated, with no evident political or ideological signalling.
How n0n gets in: credential stuffing, not zero-days
SOCRadar found no exploit payloads or CVE references in the operator workspace it recovered from an exposed staging server. Every documented compromise relied on weak or stolen credentials, through ordinary access paths:
- Credential stuffing against FortiGate SSL-VPN appliances through residential proxy pools, against an indexed set of 305 devices with a 15-thread routine and a distinctive
headerCookie: APBSID=n0n. - RDP and WinRM sweeps for default or weak passwords, driven by OCR-assisted scripts.
- Long-lived AWS IAM keys with list, read and delete rights.
- Stored administrative browser sessions for hosting-provider management, billing and monitoring portals.
- API tokens and cloud credentials harvested from source code and CI/CD pipelines.
After access, the operator created hidden administrator accounts disguised as services, such as svc_upd and svc_reg, reset root through GRUB init=/bin/bash on OCR-driven consoles, ran a gost SOCKS5 proxy as a scheduled task, and started OpenSSH on port 2222. Exfiltration ran through dedicated per-victim pipelines with byte budgets from 1 GB to 110 GB per cloud bucket, ending in Base64-encoded ZIP POSTs, SigV4 requests over Tor, and SSH-tunnelled TAR archives. The group harvests browser passwords, session cookies, mailboxes and MFA codes, and keeps persistence through hidden administrator accounts, scheduled tasks, SSH keys and proxy tools.
None of this needs a zero-day. It needs password-only remote access, over-privileged cloud keys and administrative sessions that can be stolen from a browser, and all three still exist in plenty of environments.
n0n’s encryptor: AES-256-CBC with a reused hardcoded key
SOCRadar recovered the original C++ source of the locker from the staging server. It is an AES-256-CBC file encryptor with no network capability, no key exchange, no per-victim key derivation, no ransom-note logic, no persistence and no anti-analysis. External wrapper scripts handle everything else, including stopping services, disabling security software and deleting shadow copies.
Each build uses a single hardcoded 32-byte key. The nenc.cpp build and the PowerShell encryptor used against FinSoft share an identical key, so a decryptor obtained by one victim would likely unlock the other’s files. SOCRadar’s assessment is that the paid “decryptors” are that static key.
That does not make every n0n victim recoverable. The wrapper scripts can destroy backups and shadow copies before or during encryption, and the Windows wrapper runs vssadmin delete shadows /all /quiet, wipes free space with cipher /w, and reruns encryption every five minutes as a SYSTEM scheduled task. Still, the cryptographic design is unsophisticated, and the key reuse looks like a shortcut from a group optimising for deployment speed over resilience against recovery.
AI-assisted ransomware: what the opencode workflow shows
SOCRadar found the operator working with opencode, an open-source AI coding agent, configured with 154 model entries across seven providers. These included commercial gateways, “uncensored” and “abliterated” models with safety training removed, and a self-hosted Qwen 27B model on a rented GPU server. Billing records showed more than $7,216 in API spending as of 22 October 2026, concentrated on the safety-removed models. The agent produced hundreds of numbered, iteratively patched scripts, stored encryption keys in its own memory file, and drove the command-and-control panel through saved session cookies.
SOCRadar describes a human-directed operation supported by AI, not a fully autonomous attack system. The AI sits behind the door, and the entry vector was still a password. That cuts against the hype about autonomous AI ransomware campaigns and against the claim that AI is irrelevant to offensive work. n0n shows a middle case: AI lowers the marginal cost of producing and repairing attack scripts, but an attacker still needs a credential, an exposed VPN or an over-privileged cloud key to get in.
n0n’s extortion model and Bitcoin payment trail
The extortion model has moved toward leaving victims less room to act. SOCRadar describes three stages: exfiltrate and encrypt or destroy while the incident stays private; list the victim on the leak site to build pressure through media and search visibility; then publish archives through ZIP downloads and web-seeded torrents.
In the October wave, n0n dropped negotiation entirely. Victims had to pay the listed address in full or face automatic publication when the countdown expired: 96 hours for mid-October listings and seven days for a maritime batch. Earlier, the group charged a 10% upfront fee just to access its negotiation portal, which stopped victims from inspecting the stolen files before paying.
The payment trail is thin. n0n is Bitcoin-only and derives a unique address per victim from a single BIP84 seed. SOCRadar collected 55 addresses and found no transactions, balances or pending mempool activity on any of them as of 16 October 2026. That does not prove no ransoms were paid, since payments could run through other addresses or intermediaries. It does fit a group optimising for public pressure and fast victim turnover over long negotiation cycles.
How to defend against n0n’s access paths
The defensive advice holds whether or not the Visa, Mastercard and Chase claims are ever verified. The paths n0n is documented to use are internet-facing FortiGate SSL-VPN without phishing-resistant MFA, RDP and WinRM exposed with weak or default credentials, long-lived AWS IAM keys with list, read and delete rights, administrative browser sessions stored on shared workstations, and API tokens committed to source code or CI/CD pipelines.
Controls that would have blocked every documented intrusion:
- Phishing-resistant MFA on remote access and cloud consoles.
- Short-lived role assumption in place of long-lived keys.
- Network-level restriction of RDP and WinRM.
- Separation of administrative sessions from general-purpose browsing.
Behaviours worth detecting:
- Hidden administrator accounts with service-like names.
- Scheduled tasks running proxy tools or SSH on non-standard ports.
vssadminshadow-copy deletion.- Large outbound transfers to cloud storage or public file-hosting services.
The AI-agent configuration is not visible on a victim network, but the scripts it produces are. The iterative patching pattern, with hundreds of numbered variants, may leave artefacts that differ from human-written tooling in volume and naming discipline.
What Visa, Mastercard and Chase customers should do now
For cardholders, the practical risk is targeted phishing and account takeover that uses leaked or claimed data to look authentic. Card fraud from a payment-card dump is the lesser concern, given what the listings describe. The sensible response is the ordinary one, applied with more attention than usual while claims and service disruption are in the news:
- Treat unsolicited messages about the breach, account lockouts, failed transactions or refunds as suspect.
- Verify through a known channel instead of a link or number in the message.
- Do not reuse passwords across financial and non-financial accounts.
Organisations that use payment-network programme, marketing or analytics vendors in the Asia-Pacific region should check what cardmember data those vendors hold, review external access and integration credentials, and prepare for phishing that exploits the claims even if the claims turn out to be exaggerated.
Frequently asked questions about the n0n claims
Did n0n breach Visa, Mastercard or JPMorgan Chase?
The available evidence does not establish that. Visa confirmed an incident at a third-party provider and said its own systems were not impacted. Chase attributed a same-day outage to a software bug and has not confirmed the data claim. Mastercard has confirmed nothing.
Were payment-card numbers exposed?
No payment-card numbers have been confirmed exposed. The claimed datasets are marketing and cardmember-programme records, and Visa said there is no indication that card numbers were affected.
Was the JPMorgan Chase outage caused by n0n?
Chase told Reuters the outage came from a software bug in on-premise hardware systems and that there was no evidence it was cyber-related. The listing appeared about two hours and forty-four minutes before Downdetector began recording reports, but that timing proves no link.
How does the n0n ransomware group get into networks?
Through weak or stolen credentials. SOCRadar documented credential stuffing against FortiGate SSL-VPN appliances, RDP and WinRM password sweeps, long-lived AWS IAM keys, stolen administrative browser sessions, and API tokens taken from source code and CI/CD pipelines. It found no exploit payloads or CVE references.
Can n0n victims decrypt their files with a shared key?
Possibly. SOCRadar found that builds use a single hardcoded key and that two builds share the same one, so a decryptor from one victim would likely work for another. Recovery is still not guaranteed, because wrapper scripts can destroy backups and shadow copies first.
What the evidence establishes
A group with a credential-based access model, lightweight encryption with a key-reuse flaw and an AI-assisted scripting workflow listed three of the most recognisable names in financial services and set a 48-hour clock. Visa confirmed an incident at a third-party provider and denied impact to its own systems. Chase attributed a same-day outage to a software bug and did not confirm the data claim. Mastercard said nothing.
The evidence does not establish that n0n breached Visa, Mastercard or JPMorgan Chase. It does establish that the group’s operating model is real, that the access paths it uses are common, and that the third-party and marketing-data layer around payment networks is a live exposure surface. My position is to neither dismiss the claims for being unverified nor treat them as proof of a payment-system compromise. Close the credential paths, shrink the blast radius of marketing and programme data, and expect the phishing wave that follows the headlines.







