CISA Adds Seven Actively Exploited Vulnerabilities to KEV Catalog, Including SonicWall and LiteLLM Flaws Targeting AI Infrastructure
CISA has added seven vulnerabilities confirmed to be actively exploited in the wild to its Known Exploited Vulnerabilities (KEV) catalog. Two of the flaws carry maximum CVSS scores of 10.0, and several affect AI infrastructure for the first time.
The catalog update includes a critical Server-Side Request Forgery vulnerability in SonicWall SMA 1000 series devices (CVE-2026-83548, CVSS 10.0) that allows unauthenticated remote access to internal sensitive functions. A companion command injection flaw (CVE-2026-83549) enables attackers who obtain administrative privileges to execute arbitrary system commands, forming a complete attack chain.
Another high-severity issue affects JFrog Artifactory (CVE-2026-82329, CVSS 9.8). Under default configurations, network-reachable attackers can obtain full administrator rights, potentially exposing build artifacts, container images, and hardcoded credentials stored in the repository.
Additional entries cover a SQL injection vulnerability in Sangoma Switchvox (CVE-2026-9586, CVSS 9.3), a command injection flaw in the Kestra workflow engine (CVE-2026-49869, CVSS 10.0), an HTTP request smuggling issue in Starlette (CVE-2026-48710), and an authentication bypass in the LiteLLM MCP interface (CVE-2026-59822, CVSS 8.8).
Security researchers from Microsoft and Wiz report that threat actors are now systematically targeting AI infrastructure components including LiteLLM gateways, Flowise, LangChain, Ollama, ChromaDB, and MCP servers. The Qilin ransomware group has already leveraged LiteLLM vulnerability chains to bypass authentication and achieve remote code execution.
After compromising a LiteLLM gateway, attackers query the PostgreSQL database tables LiteLLM_ProxyModelTable and LiteLLM_VerificationToken to exfiltrate model configurations and API keys in bulk. They frequently deploy XMRig cryptocurrency miners and terminate competing mining processes before exiting.
Similar intrusions against Kestra instances have involved establishing reverse shells, enumerating Docker environments, deploying miners, and exfiltrating data through the platform’s own key-value interfaces to minimize filesystem artifacts.
CISA has instructed federal agencies to complete remediation by September 5, giving organizations roughly 48 hours. Recommended actions include immediate asset discovery for all listed products, incorporation of AI gateways into formal asset inventories, rotation of all exposed model API keys, and monitoring for unauthorized mining processes and changes to authorized_keys files.
Related articles
Gmail Address Variants Bypass String Comparison and Enable Multiple Registrations
A single user created four separate accounts using valid Gmail address formats that all deliver to the same inbox: Ivan.Petrov@gmail.com, ivanpetrov@gmail.com, ivan.petrov+shop@gmail.com, and IVAN.PETROV@gmail.com. Simple string comparison in the registration code treated these as distinct entries, allowing circumvention of new-user promotions, free-tier limits, and one-application-per-person rules. The root cause lies in three Gmail-specific behaviors: case-insensitive local parts, plus-addressing that ignores everything after the + symbol, and dots that carry no semantic meaning in the local part. Additional risks arise from IDNA homograph domains that visually mimic legitimate addresses but encode to different Punycode strings. Proper mitigation requires storing both the original email for delivery and a domain-aware normalized form for uniqueness checks, combined with explicit IDNA validation on registration.
Dangling DNS Records Expose Companies to Subdomain Takeovers Years After Projects Shut Down
A marketing team requested reactivation of a three-year-old promotional landing page, only to discover that its domain now served an unrelated third-party website. The root cause was a forgotten CNAME record in the company's DNS zone that continued pointing to a project name on external hosting after the original service had been decommissioned. Because many hosting providers use wildcard DNS entries, standard availability checks and simple DNS resolution tests fail to detect that the target project no longer exists. Attackers who re-register the freed project name can then serve content under the company's legitimate subdomain, automatically obtaining valid TLS certificates and inheriting domain-wide cookies. The article details how such dangling records accumulate across infrastructure, marketing, and acquired company zones, and explains the concrete risks including cookie theft, bypass of Content-Security-Policy, and unauthorized domain ownership proofs. It concludes with a practical remediation workflow: export the full zone, assign owners to every record, validate external targets by application response rather than DNS resolution, and embed record removal into decommissioning procedures.
Google Patches Actively Exploited V8 Zero-Day in Chrome (CVE-2026-85046)
Google has released an urgent update for Chrome to address CVE-2026-85046, a high-severity zero-day vulnerability in the V8 JavaScript engine that is being actively exploited in the wild. The flaw is a type confusion issue within the Maglev compiler, specifically tied to optimizations around Array.prototype.sort, which can lead to arbitrary memory read and write primitives inside the JavaScript heap. Attackers can exploit the vulnerability by tricking victims into visiting a specially crafted HTML page, enabling remote code execution within the browser sandbox. The update brings Chrome to version 152.0.7977.82 or higher on Windows, macOS, and Linux, while also addressing a total of 12 vulnerabilities. The issue was reported by researcher Salvatore Gulizia, known as Serotav, who received a $1,000 reward. Other Chromium-based browsers such as Edge, Brave, Opera, and Vivaldi will require corresponding updates from their vendors.
Google Patches Actively Exploited V8 Zero-Day in Chrome (CVE-2026-85046)
Google has released an urgent update for Chrome to address CVE-2026-85046, a high-severity zero-day vulnerability in the V8 JavaScript and WebAssembly engine that is being actively exploited in the wild. The flaw is a type confusion issue tied to the Maglev compiler and optimizations around Array.prototype.sort, which can lead to arbitrary memory read and write primitives inside the JavaScript heap. Successful exploitation allows a remote attacker to execute arbitrary code within the browser sandbox after a victim visits a specially crafted HTML page, a technique commonly seen in malvertising and compromised-site campaigns. The fix is included in Chrome 152.0.7977.82 and later for Windows, macOS, and Linux, and the update also resolves a total of twelve vulnerabilities. The vulnerability was reported on 4 August 2026 by researcher Salvatore Gulizia, known as Serotav, who received a $1,000 reward. Other Chromium-based browsers such as Microsoft Edge, Brave, Opera, and Vivaldi will require corresponding updates from their respective vendors.