cybr.cx Daily Digest — October 06, 2026

Critical Vulnerabilities

⚠️ Actively exploited — CVE-2026-102489 & CVE-2026-102490 | Zammad | No CVSS in NVD feed
These two Zammad vulnerabilities are being chained in the wild right now, and the remediation deadline has already passed (October 5). CVE-2026-102489 is a session fixation flaw that allows unauthenticated remote attackers to gain code execution as the zammad system user; CVE-2026-102490 then escalates that foothold to full root via improper privilege management. If your organisation runs Zammad for ticketing or support operations, treat this as an emergency patch — a complete system compromise from an unauthenticated starting position is as bad as it gets.

⚠️ Actively exploited — CVE-2026-104286 | Fortinet FortiMail | No CVSS in NVD feed
An unauthenticated path traversal combined with a NULL byte injection flaw lets attackers write arbitrary files to the underlying OS via crafted HTTP/S requests — the classic recipe for webshell deployment. CISA's remediation deadline was October 4, meaning unpatched systems have had an open window for two days. FortiMail sits at the perimeter and processes untrusted input at volume; prioritise this immediately.

⚠️ Actively exploited — CVE-2026-76504 | Cisco Catalyst SD-WAN Manager | No CVSS in NVD feed
A hex encoding quirk in URI handling allows unauthenticated remote attackers to access SD-WAN Manager with full admin privileges. Exploitation requires nothing more than a crafted HTTP request. SD-WAN management planes are high-value targets — compromise here means an adversary can reshape your entire WAN topology.

⚠️ Actively exploited — CVE-2026-88779 | Citrix NetScaler ADC / Gateway | No CVSS in NVD feed
An out-of-bounds memory buffer vulnerability that can be triggered to cause denial of service — and in this product family, DoS at the ADC layer typically means authentication bypass conditions follow. Citrix NetScaler has been a persistent attacker favourite; CISA's due date was yesterday (October 7). Patch or isolate now.

⚠️ Actively exploited — CVE-2026-86950 | Apple iOS, macOS, iPadOS — CoreGraphics | No CVSS in NVD feed
An out-of-bounds write in CoreGraphics can lead to arbitrary code execution. Apple has shipped fixes; the concern here is unmanaged personal devices and BYOD fleets that haven't applied updates. Mobile device management policies should be enforcing patch compliance checks immediately.

CVE-2026-101919 | Red Hat HyperShift Operator | CVSS 8.8 — HIGH
The HyperShift operator copies user-supplied kubeconfig secrets directly into privileged control plane namespaces without sanitisation. An authenticated user with cluster and secret creation rights can embed malicious executable plugins that run with control plane privilege. In multi-tenant Kubernetes environments this is a significant lateral movement and privilege escalation risk — review RBAC permissions around secret creation immediately.

CVE-2026-105221 | gist RubyGem (< 6.1.0) | CVSS 7.4 — HIGH
The library sets VERIFY_NONE for TLS connections, meaning any on-path attacker can present a rogue certificate and intercept GitHub API traffic in cleartext. OAuth tokens and credentials transiting this gem are fully exposed. Update to 6.1.0 or later; audit CI/CD pipelines that use this gem for token leakage.

CVE-2026-105294 | Legcord 1.1.0–1.3.0 | CVSS 7.4 — HIGH
A configuration injection flaw lets JavaScript executing in a Discord page write arbitrary config keys via the window.legcord settings bridge. An attacker exploiting a Discord XSS can persistently set --proxy-server and --ignore-certificate-errors Chromium flags, routing all of the client's traffic through an interception proxy across reboots. Users of this third-party Discord client should upgrade immediately.

Additional SQL injection CVEs (CVE-2026-105166, CVE-2026-105170, CVE-2026-105182, CVE-2026-105232) affecting small open-source PHP applications (food-waste-management-system, Online Reviewer Management System) and an SSRF in feelcrm-os (CVE-2026-105290) round out today's feed — low real-world exposure but public exploits are available.

Headline News

Denmark's CPR Registry Breach Exposes 8.8 Million People
Denmark's Civil Registration System — the CPR — has suffered an unauthorised access incident affecting the personal data of approximately 8.8 million individuals, which is essentially the entire Danish population plus diaspora registrants. The CPR database is a foundational identity infrastructure, linking health records, tax filings, banking, and government services to a single national identifier, making this breach qualitatively different from a typical credential dump. For practitioners, this is a reminder that centralised identity registries represent catastrophic single points of failure: when they go, everything downstream is affected simultaneously. The incident will almost certainly trigger cascading fraud and social engineering campaigns as threat actors cross-reference CPR numbers with other leaked datasets. Organisations operating in Denmark or handling Danish customer data should expect elevated phishing and identity fraud attempts in the coming weeks.

Chinese APT Impersonates Anthropic Employee in AI Espionage Campaign
A Chinese state-linked threat actor has been observed conducting targeted social engineering against US artificial intelligence policy experts by impersonating employees of leading AI labs, including Anthropic. The campaign's goal appears to be extracting strategic intelligence about AI capabilities, safety research timelines, and policy positioning — the kind of information that informs national AI competitiveness decisions rather than providing direct technical theft. This mirrors well-documented patterns from groups like APT10 and APT41, which have previously used spearphishing and persona construction to penetrate research communities. For security teams defending AI organisations, this underscores that the attack surface is not just the model infrastructure but the human networks of researchers, policy advisors, and executives who discuss sensitive roadmap details over email and messaging platforms. Personnel security awareness programmes at AI labs need to specifically address vendor and peer impersonation scenarios.

ShinyHunters-Linked Suspect Detained in Jordan Following FBI Recruitment Portal Breach
A teenager known online as "Rey," linked to the ShinyHunters group, has been detained in Jordan in connection with the breach of the FBI's job recruitment portal disclosed last month. ShinyHunters claimed to have exfiltrated contractor and applicant data from the portal, a breach with obvious counterintelligence implications given the sensitivity of individuals who apply for federal law enforcement positions. FBI Director Kash Patel has indicated further arrests are being pursued, suggesting law enforcement has visibility into more of the group's membership. ShinyHunters has a long track record of high-profile intrusions — Ticketmaster, AT&T, and others — typically leveraging credential stuffing and third-party cloud misconfigurations rather than novel exploits. The Jordan detention is notable because it demonstrates continued international cooperation on cybercrime extradition, even for suspects who may have believed they were in jurisdictions with limited exposure.

Schrödinger's Feed

A new post-quantum content verification protocol — dubbed TIP — has been publicly launched, designed to cryptographically authenticate digital content using PQC primitives, with an eye toward combating AI-generated disinformation at scale. The protocol was unveiled at Berlin's AI, Data & Quantum Summit, where the convergence of quantum computing and AI was a dominant theme among over 2,500 attendees. Meanwhile, Germany's federal government is committing €640 million to develop fault-tolerant quantum processing units by 2030 across six competing consortia spanning neutral atom, trapped ion, and superconducting architectures. Practitioners should note the narrowing timeline: fault-tolerant quantum at scale by 2030 means the cryptographic migration window is shorter than many organisations' current patch cycles — if your PKI roadmap doesn't include PQC transition planning, it's already behind.

/dev/random

Google has frozen its open-source bug bounty programme through 2027 after being overwhelmed by a flood of AI-generated vulnerability reports — the submissions were, in the parlance of the field, "hallucinations all the way down." Maintainers reportedly found themselves spending more time debunking plausible-sounding but entirely fictional CVEs than reviewing real research. It's a genuinely novel failure mode: the bug bounty programme, a mechanism designed to surface real flaws, is being DoS'd by synthetic ones. Somewhere, a penetration tester is drafting a report titled "CVE-2026-CHATGPT: AI Model Allows Unauthenticated Waste of Everyone's Time."