🔴 PaperCut zero-days hit CISA KEV with 800+ servers exposed, 🎯 Exchange PoC amid Oct EOL deadline, 🔥 Fire Ant plants GRE tunnels on Cisco IOS XR, 🕵️ Fake AI crawlers scan for .env files, 🌐 BGP hijack delivers malicious Virtualizor updates, 🧟 SLEEPWALKER passive backdoor, ⚡ McKesson ShinyHunters extortion

[UPDATE] PaperCut NG/MF exploitation has escalated from emergency patches to a confirmed CISA KEV listing, with both CVE-2026-81578 and CVE-2026-82078 added to the catalog on September 1. Defused honeypots show attackers are chaining the authentication bypass to hijack PaperCut's external user-lookup mechanism and dump database tables via Derby, diverging from the RCE path documented in public writeups. Over 800 PaperCut MF and NG servers remain exposed and vulnerable according to Shadowserver tracking, and Rapid7's Metasploit module for the exploit chain has already been merged, lowering the skill barrier for mass exploitation. This is now a patch-now-or-face-regulatory-consequences situation.

Microsoft Exchange on-premises administrators face a converging threat. A public proof-of-concept repository claims a pre-authentication RCE chain targeting Exchange via CVE-2026-62911, though Microsoft and the Zero Day Initiative classify the issue as an elevation-of-privilege capture-replay bug rather than standalone unauthenticated RCE. The PoC references Exchange's Mailbox Replication Service Proxy and claims NTLM relay as the initial access vector, but should be treated as unverified until independently validated in an isolated lab. Separately, Microsoft confirmed Extended Security Update support for Exchange 2016 and 2019 ends in October 2026 with no extensions planned, giving organizations approximately three months to migrate to Exchange Online or Exchange Subscription Edition. The convergence of a working exploit chain demonstration with an imminent end-of-support deadline makes on-premises Exchange an increasingly attractive target.

Chinese state-sponsored group Fire Ant has expanded from VMware hypervisor targeting into Cisco IOS XR router compromise, planting GRE tunnel interfaces with no matching configuration or commit history, then deploying custom malware that persists through a fake system service running only during alternating hours. The implant selectively suppresses syslog messages to hide tunnel activity and opens outbound Telnet connections to Fire Ant infrastructure. Beyond routers, Fire Ant also compromised TACACS authentication servers and Linux management hosts, using administrative access to capture router traffic and exfiltrate PCAP files to external FTP servers, potentially exposing internal topology, admin connections, authentication flows, and routing relationships.

GreyNoise researchers identified a campaign in which threat actors impersonate web crawlers from OpenAI, Anthropic, DeepSeek, Google, Perplexity, and Amazon to scan internet-facing servers for exposed secrets. The activity ran from late July through late August 2026 across 824 IP addresses, using forged user-agent strings to request .env files, AWS credential stores, private keys, and password databases. Six of the impersonated names, tied to Anthropic, OpenAI, Google, and Perplexity, appeared at nearly identical volumes, suggesting coordinated infrastructure. The campaign exploits a fundamental weakness in how websites identify bots: user-agent headers are self-reported and trivially spoofed. Organizations should validate crawler identity against known IP ranges rather than trusting user-agent strings.

Attackers hijacked BGP routing for Softaculous infrastructure, diverting traffic to deliver a malicious Virtualizor update to hypervisor servers. The incident ran from approximately 20:57 UTC on August 28 until 06:10 UTC on August 30, 2026. Using AS62390 (NexonHost), the attacker announced the more-specific prefix 162.55.80.0/24 through AS6204 (Zet), winning BGP selection wherever propagated while retaining Hetzner's ASN in the path to appear less suspicious. The diversion also enabled valid Let's Encrypt certificates for virtualizor.com and related domains, so affected clients connected over apparently valid HTTPS sessions and received the tampered package. This is a clean example of supply-chain compromise via BGP, where no software vulnerability was exploited, only the routing fabric that underpins internet trust.

SLEEPWALKER is a previously undocumented Windows backdoor that uses five distinct covert channels for command reception: raw network packets, DNS queries, SMB named pipes, ICMP ping traffic, and VMware VMCI communication. The malware impersonates Microsoft's dpapi.dll and is side-loaded by ESET Management Agent (ERAAgent.exe), remaining inert unless running inside the expected ESET process. Once activated by a precisely crafted encrypted trigger packet, it decrypts and executes attacker-supplied bytecode through a custom interpreter supporting 23 instructions. The passive trigger model is the key differentiator: SLEEPWALKER receives commands rather than requesting them, which means network-level monitoring alone will not catch it.

A GitHub user operating under the names MSNightmare and Nightmare Eclipse published a proof-of-concept exploit called HardBreacher claiming to exploit an unpatched local privilege escalation vulnerability in Kaspersky Endpoint Security for Windows version 14.0.0.504 on fully patched Windows 11 25H2. Kaspersky has not confirmed the vulnerability and no CVE has been assigned. The exploit requires prior local code execution, making it a post-compromise escalation tool. If confirmed, an attacker who has already gained a foothold could escalate from low-privilege to system-level access, bypassing the very product meant to detect and prevent such movement.

McKesson, which distributes roughly one-third of all pharmaceuticals used in North America, disclosed a cyberattack on August 29 after ShinyHunters added the company to its data-leak site. The intrusion ran from August 21 to 25 before detection, with attackers hitting third-party applications and exfiltrating data from customers in the oncology, multispecialty, and medical-surgical business units. McKesson says distribution centers remain operational, describing the impact as temporary service interruptions. ShinyHunters is known for large-scale data theft followed by extortion rather than encryption, typically imposing a countdown timer before publishing stolen data.

The Shai-Hulud Trinitite campaign compromised the npm package @7nohe/openapi-react-query-codegen, a TanStack Query code-generation library with more than 150,000 weekly downloads. The malicious releases deploy an evolved Mini Shai-Hulud worm that steals developer, cloud, CI/CD, package-registry, Kubernetes, Vault, and source-control credentials, then uses recovered access to spread through additional packages. Attackers exploited the compromised package's release workflow, which accepted a pull-request comment containing "npm publish" as a publishing trigger, giving attacker-controlled code a valid GitHub Actions OpenID Connect trusted-publishing token. Organizations using TanStack Query or depending on @7nohe/openapi-react-query-codegen should audit for malicious versions and rotate exposed credentials.

Watchlist: deploy PaperCut Emergency Patch Release 2 immediately and audit for Derby database access anomalies; validate Exchange PoC claims independently and begin migration planning before October end-of-support; audit Cisco IOS XR routers for unexplained GRE tunnel interfaces and TACACS server logs for anomalous authentication; validate AI crawler identity against known IP ranges and audit public servers for exposed .env files; monitor BGP announcements for unexpected prefix specificity changes; track McKesson for ShinyHunters' data publication timeline; audit npm dependencies for @7nohe/openapi-react-query-codegen compromised versions.