>
av.Amit VijayanCYBERSECURITY & RESEARCHLet\u2019s connect
Back to the library

Cyber news

Daily Cyber Threat Brief — November 21, 2026: Bell Canada confirms 'agentic' hacker stole enterprise customer data

Thumbnail for: Daily Cyber Threat Brief — November 21, 2026: Bell Canada confirms 'agentic' hacker stole enterprise customer data

Case file — Lead incident: Bell Canada, one of Canada's largest telecommunications providers, has confirmed that an "agentic" threat actor breached its systems on November 17, 2026, and stole enterprise customer data. The attacker reached a pre-production environment of Bell's Maximo system, the platform used to order and provision telecommunications circuits, and took business contact information — names, job titles, corporate email addresses, phone numbers and business addresses — along with ticket titles, some IP addresses tied to network elements, and limited contextual information. Bell says the stolen data did not include credentials, highly sensitive personal information, PDFs, file attachments or other supplementary information stored in Maximo. The company has not said how many enterprise customers were affected, how much data was taken, or how the attacker got in, and it has not identified the actor: "We currently do not know the attribution," Bell told BleepingComputer. Bell is sharing indicators of compromise and threat-hunting information with affected enterprise customers and the Canadian Centre for Cyber Security, but has not released them publicly. Status: confirmed breach; investigation ongoing.

Incident timeline

DateDevelopment
Early September 2026Suspected China-nexus group Armory Castle exploits a vCenter zero-day, CVE-2026-123890, inside a network it had already breached, using a custom tool dubbed BONESAW to create an SSO administrator account and steal the server's directory database.
September 8–10, 2026AI agents running the open-source Cairn framework find a previously unknown path-traversal flaw in the MONAI Label medical imaging application and use it to take root control of a server at East Asian educational institutions.
November 16, 2026N-able begins investigating a service disruption affecting its Mail Assure and SpamExperts Hosted Cloud services across the Americas, Asia-Pacific and Europe.
November 17, 2026Bell Canada is breached: an "agentic" threat actor accesses systems including the Maximo circuit-ordering platform and exfiltrates enterprise customer data.
November 18, 2026N-able confirms the disruption was caused by a security incident and disables the Mail Assure and SpamExperts management portal and APIs. The MONAI Label project publishes a security advisory for the flaw Cairn exploited.
November 19, 2026Broadcom patches the vCenter zero-day in VMSA-2026-0009, fixing CVE-2026-123890 in vCenter 8.0 U3m, 9.0.2.0200 and 9.1.1.0200, with no workaround offered.
November 20, 2026Bell publicly confirms the breach; Amazon Threat Intelligence publishes its report on the Cairn campaign; BleepingComputer reports the N-able breach; N-able says remediation is progressing but portal and API access remain disabled.

1. Bell Canada confirms "agentic" hacker stole enterprise customer data

Canada's largest telecom has confirmed a breach with an unusual label attached. Bell Canada told BleepingComputer that an "agentic" threat actor — language suggesting an attacker using AI-powered tooling — breached its systems on November 17, 2026, and stole enterprise customer data. The company has not explained what made the attacker "agentic", nor how it got in.

The intrusion came to light after a Bell customer shared a breach notification from Bell's Information Security Team with BleepingComputer. According to that notification, the unauthorised party accessed a pre-production environment of Maximo, the system Bell uses to order and provision telecommunications network connections. Bell later told BleepingComputer the actor reached beyond that one system. "Bell is investigating a cybersecurity incident that occurred on November 17 in which an agentic threat actor accessed some of our systems, including a system used for ordering and provisioning circuits, and exfiltrated some enterprise customer data," the company said. "Upon learning of the breach, we immediately contained the incident and secured our systems. There has been no impact to customer services."

What was taken is business data, not consumer records. The exposed information includes business contact details — names, job titles, corporate email addresses, phone numbers and business addresses — plus information contained in ticket titles, some IP addresses associated with network elements or systems, and limited contextual information. Bell was explicit about what was not in the dataset: no credentials, no highly sensitive personal information, and none of the PDFs, file attachments or other supplementary information stored in Maximo. Bell has not disclosed how many enterprise customers were affected or the total amount of data stolen, and it has not identified the attacker. The company says it is sharing indicators of compromise and threat-hunting information with enterprise customers and the Canadian Centre for Cyber Security, but it has not released the indicators publicly and did not share them when BleepingComputer asked. Bell has also contacted the RCMP's cybercrime unit, and says nobody has contacted the company claiming responsibility.

Investigation notes

What is confirmed: the November 17 breach date; the pre-production Maximo environment as an entry point; the categories of enterprise data taken and the categories Bell says were not taken; the company's statement that it contained the incident with no impact to customer services; and the sharing of indicators with customers and the Canadian Centre for Cyber Security.

What is not known: how the attacker got in, how long it had access, whether production systems were reached as well as the pre-production environment, how many customers were affected, how much data was taken, and who the attacker is. Bell has not said what specifically makes the actor "agentic".

What to watch: whether Bell publishes indicators of compromise, which would let other organisations check for related activity; any disclosure of the intrusion method and the affected-customer count; and whether the stolen business contact data is used in social-engineering attempts impersonating Bell — the company's own notice warns recipients to be wary of exactly that.

2. VMware patches vCenter zero-day exploited since September by Armory Castle

Broadcom has patched a critical VMware vCenter Server authentication bypass that a suspected Chinese state-sponsored group exploited as a zero-day for more than two months. The vulnerability, tracked as CVE-2026-123890 and rated 9.8 on the CVSS scale, sits in vCenter's VMware Directory Service. Broadcom's advisory, VMSA-2026-0009, published on November 19, 2026, says an attacker with network access to vCenter can bypass authentication; the company lists no workaround.

Google's Mandiant and the Google Threat Intelligence Group, who reported the flaw to Broadcom, documented the exploitation. They say the suspected China-nexus espionage group tracked as Armory Castle (formerly UNC5221) exploited CVE-2026-123890 in early September 2026, as part of the same intrusion in which the group first compromised a BeyondTrust Remote Support appliance through a separate zero-day, CVE-2026-102764, then moved laterally through internal Linux appliances that had no endpoint detection before scanning for VMware ports. The group used a custom Golang utility, BONESAW, to exploit the vCenter flaw: Mandiant observed it creating a vCenter Single Sign-On administrator account and stealing the server's directory database, which can be used to extract further credentials. The tool can also exfiltrate the SAML Security Token Service signing key and forge administrative SAML assertions, and the attackers extracted the vCenter master symmetric key and PostgreSQL database credentials. Broadcom itself has not confirmed the exploitation; Mandiant and GTIG documented it.

The flaw is an unauthenticated Secure Remote Password (SRP-6a) session key disclosure vulnerability over DCE/RPC in the vmdird service, exposed on TCP port 2012. Successful exploitation produces a remote authentication bypass that creates an authenticated session as [email protected]. It affects vCenter 8.0, 9.0 and 9.1 and the vCenter builds shipped inside VMware Cloud Foundation, vSphere Foundation and the Telco Cloud products. The fixes are vCenter 8.0 U3m, 9.0.2.0200 and 9.1.1.0200, plus VCF 5.2.5. The same advisory covers four further flaws in SDDC Manager and VMware Tools for Windows. One caution for defenders: applying the patch does not invalidate a stolen SAML signing key, so any environment that may have been hit needs credential and key rotation, not just the update.

Investigation notes

What is confirmed: the CVE-2026-123890 identifier and 9.8 CVSS rating; Broadcom's November 19 patch in VMSA-2026-0009 with the fixed versions listed; Mandiant and GTIG's account of exploitation in early September by Armory Castle, including the BONESAW tool, the SSO administrator account creation and the directory database theft; and the absence of a workaround.

What is not known: how many organisations were hit — the public reporting describes one intrusion, and Broadcom has not confirmed exploitation at all. The victim organisation in the Mandiant report has not been named.

What to watch: whether CISA adds CVE-2026-123890 to the Known Exploited Vulnerabilities catalog, which would set a federal remediation deadline; vCenter logs for the indicators in the Mandiant report, especially unexpected SSO administrator accounts and access to TCP port 2012; and rotation of SAML signing keys and directory credentials in any environment that was exposed.

3. N-able breach knocks out Mail Assure and SpamExperts management worldwide

N-able has confirmed that attackers who breached one of its servers caused a major outage affecting its Mail Assure and SpamExperts Hosted Cloud services across the Americas, Asia-Pacific and Europe. The company first told customers on November 16, 2026, about a service disruption it attributed to a combination of platform configuration and service dependencies. On November 18, after further investigation, it determined the disruption was caused by a security incident: unauthorised access to a server in the Mail Assure and SpamExperts Hosted Cloud environment.

The company disabled the management portal and APIs as a precaution, along with quarantine, logging and mailbox synchronisation services. Customers cannot sign in to the control panel, view or release quarantined messages, review logs, or make configuration changes such as adding or removing domains while those services are down. Core mail flow and inbound and outbound filtering have continued to operate throughout. N-able says it has found no evidence so far that customer data was compromised, including access to tenants' mailboxes, and that the investigation is ongoing. In a status update at 08:06 UTC on November 20, the company said remediation had progressed overnight and engineering teams were addressing outstanding items and testing completed measures, with portal and API access still disabled as a precaution. N-able stressed the incident is unrelated to the separate N-central vulnerabilities known as Able2Shell, 4EverAble2Shell and Agent2Shell.

Investigation notes

What is confirmed: the November 16 disruption; the November 18 determination that a security incident caused it; unauthorised access to one hosted-cloud server; the management portal, APIs, quarantine and logging remaining offline; continued mail flow and filtering; and the absence, so far, of evidence of customer data compromise.

What is not known: how the attackers got in, whether any information was removed, and when the management services will be restored. No attacker has been named and no intrusion method has been disclosed.

What to watch: N-able's status page for a restoration date, any revision to the data-compromise assessment, and whether the company discloses an initial-access vector — Mail Assure and SpamExperts sit in front of customer mail, so the access path matters to every tenant.

4. AI agents found a zero-day on their own and took root on a university server

Threat actors used an autonomous AI agent platform to discover a previously unknown vulnerability and exploit it to take complete control of a server — with the human operator giving only a high-level instruction. Amazon Threat Intelligence published the findings on November 20, 2026. A suspected Chinese-nexus actor pointed the open-source framework Cairn, a penetration-testing tool that coordinates multiple AI agents, at educational institutions in East Asia outside mainland China between September 8 and 10, 2026, and simply instructed the agents to obtain shell access.

The agents mapped the target application's API, found a path-traversal zero-day in MONAI Label — an open-source medical imaging application that uses AI to help researchers label medical images — and went from mapping the API to writing files to arbitrary locations on the server in about 11 minutes. In a subsequent session they wrote a Python reverse shell into sitecustomize.py, a file that loads automatically when Python starts, then triggered an API endpoint to establish root access. The flaw is now tracked as CVE-2026-133968 and affects MONAI Label 0.8.5 and earlier. Amazon reported it to the maintainers, who published a security advisory on November 18; the fix is merged and slated for the next release, with no patched version published yet. Amazon uncovered the campaign because the attackers left their Cairn instance publicly accessible without authentication, exposing their operations — including the prompts, written in Simplified Chinese, and script timestamps configured to China Standard Time, on which Amazon bases its moderate-confidence attribution. The company also published two IP addresses observed in the campaign, 142.4.98.145 and 142.4.107.34.

Investigation notes

What is confirmed: Amazon Threat Intelligence's November 20 report; the September 8–10 campaign window; the Cairn framework as the attack tool; the MONAI Label path-traversal flaw, CVE-2026-133968, affecting version 0.8.5 and earlier; the 11-minute path from API mapping to arbitrary file write; the root shell via sitecustomize.py; and the November 18 project advisory.

What is not known: how many institutions were compromised beyond the servers Amazon observed, and whether the exposed Cairn instance captured the full campaign. Attribution to a Chinese-nexus actor is assessed with moderate confidence, not confirmed.

What to watch: the release of a patched MONAI Label version; any MONAI Label instance exposed to untrusted networks should be taken offline or isolated until then, since the application runs with authentication disabled by default; and the two published IP addresses for hunting in environments that ran MONAI Label in September.

On the watch list

The Gentlemen ransomware group's claim against Nvidia remains unproven, with no sample files published and no confirmation from the company, and its countdown runs to about November 28. On the exposure front, Shodan currently shows about 786,000 exposed NetScaler instances, 505,000 exposed FortiGate appliances, 118,000 exposed Exchange servers and 12,000 exposed Danfoss devices worldwide.

Sources

Keep exploring.

All articles