VPN provider Surfshark disclosed that attackers accessed an internal engineering test server that a configuration error had left reachable from the internet, along with a proxy server used for content optimization. Surfshark says the exposure was limited to a non-production environment and included service configurations, build-related credentials, system binaries, and portions of code history, but did not reach customer data, VPN traffic, encryption keys, or production systems. The company detected the activity on August 31, contained it by September 2, rotated credentials, revoked tokens, and completed remediation within days. It is a reminder that misconfigured internet-exposed test environments remain a common and avoidable breach path, even at security-focused companies.
CISA added several actively exploited flaws in internet-facing security appliances to its catalog, ordering federal agencies to patch by September 12. The most severe, CVE-2026-20079 scored 10.0, is an authentication bypass in Cisco Secure Firewall Management Center that lets an unauthenticated attacker run scripts and gain root on the device; Cisco confirmed exploitation since August. A Citrix NetScaler authentication bypass, CVE-2026-19490, saw a surge of attacks on September 8, and a Fortinet FortiOS flaw, CVE-2025-25249, is being used to deliver a remote access trojan. Separately, CISA warned that a critical WatchGuard Firebox firewall flaw is now being exploited in ransomware attacks. Edge appliances remain prime targets.
Researchers at Wiz found that nearly one in ten internet-facing LiteLLM servers still accept "sk-1234," the example administrator key printed in LiteLLM's own setup guide. LiteLLM is an open-source AI gateway that sits between an organization's apps and the model providers it pays for, and that admin key unlocks every stored provider API key; in Wiz's tests it even reached the cloud identity credentials of the host machine. The finding accompanies a cluster of exploited LiteLLM flaws that attackers have used to run code, steal secrets, and deploy crypto miners, with one ransomware group and a Microsoft-documented breach among them. Microsoft's advice is to treat AI gateways as top-tier secrets stores.
cPanel patched a critical flaw that lets an ordinary hosting account with mail privileges take root control of the whole server. Tracked as CVE-2026-67401 and scored 9.9, it is a SQL injection in the EmailTrack mail-tracking feature that lets an authenticated account create arbitrary files and escalate to code execution as root. It affects all supported cPanel and WHM versions. On a shared server, a single cheap hosting plan or one stolen webmail password can lead to full server takeover, exposing every other tenant's sites, databases, and data. It is the third cPanel flaw since late July that turns one authenticated tenant into root, and cPanel published no indicators to hunt for.
Alby warned of a critical flaw in older versions of Alby Hub, a self-hosted Bitcoin Lightning wallet that people run on their own computer or server to hold their funds. An attacker who could reach the wallet's management interface over the internet could gain access without permission and send the owner's funds. The flaw affects versions 1.7.0 through 1.18.5, released before August 2025, and was fixed in 1.19.0 and later, with 1.24.0 the current release. Alby says one user has been affected so far and is withholding technical details for now. The core lesson is to never expose a self-hosted wallet's control interface to the public internet.
Researchers at VulnCheck found a flaw in the DeepSeek Harness, a tool that runs an AI agent's commands inside an operating-system sandbox so an agent handling untrusted files cannot write outside its workspace. Through an authentication bypass using a spoofed host header, an attacker needing no credentials or API key can call the tool's own web interface to invoke privileged commands with full-access permissions, raise the session's approval policy to unrestricted execution, and read every stored conversation. In effect, the sandbox meant to contain the agent can be switched off from outside. It is a reminder that an AI agent's isolation is only as strong as the authentication protecting its control interface.
AdaptHealth, a US network of more than 680 medical-equipment facilities, confirmed that a breach attributed to the ShinyHunters group exposed the personal, health, and insurance information of about 4.1 million people. The attackers got in by socially engineering a third-party contractor's privileged account, then reached AdaptHealth's cloud business applications, patient-management systems, and electronic health record portals, and stole a password file tied to insurance billing. It fits ShinyHunters' pattern of tricking a person into handing over access to connected cloud services, and it is the latest in a wave of large healthcare breaches this year alongside Aesto, CareCloud, and McKesson. Social security and financial data were reportedly not taken.
Microsoft shipped its largest-ever Patch Tuesday, fixing a record 974 vulnerabilities, including two Windows zero-days already exploited in attacks. Both zero-days are local privilege-escalation flaws that let an attacker gain SYSTEM access: CVE-2026-85880 is a heap buffer overflow in the Advanced Local Procedure Call component that can let code in a low-privilege sandbox escape and elevate, and CVE-2026-81963 is a link-following flaw in the Windows Update Stack. The release also includes about 20 potentially wormable flaws, remotely exploitable without authentication, across services like DNS, DHCP, SMB, and Active Directory, plus critical fixes in Exchange, SharePoint, SQL Server, and Kerberos. The sheer volume makes prioritization essential.
SAP patched a critical flaw in its kernel, tracked as CVE-2026-44756 and dubbed OVERPASS with a top score of 10.0, that lets an unauthenticated, remote attacker run commands with administrative privileges and fully compromise a system. The memory-corruption bug is in the Extended Passport processing library and is reachable over several SAP communication protocols, including through the internet-facing Internet Communication Manager, which researchers say exposes more than 10,000 SAP systems online. In the same update SAP fixed a second 10.0 flaw, a missing-authentication issue in the NetWeaver Message Server that lets attackers run code across an entire SAP cluster without credentials. Both need prompt patching.
Researchers at Sophos and ESET found attackers breaching F5 BIG-IP APM access gateways and installing a stealthy Linux rootkit that ESET calls PoisonedRefresh. Rather than dropping a file on disk, it hides a web shell in memory, hooks into the Apache and PHP components, tampers with SELinux settings, and uses disguised requests to run commands while blending into normal traffic. Crucially, it persists across device upgrades, so patching the appliance alone does not remove it. The intrusions likely began by exploiting a critical remote code execution flaw that F5 had earlier downgraded to a mere denial-of-service issue, which may have led some organizations to deprioritize patching it.