CISA added five flaws exploited by the China-linked actor Flax Typhoon to its Known Exploited Vulnerabilities catalog, including CVE-2015-3306 (CVSS 10.0), an improper access control flaw in ProFTPD, and set an October 11 patch deadline for federal agencies.
Researchers published a full working exploit for a pre-authentication remote code execution flaw in AnyDesk for Linux that grants root access before any connection is approved. AnyDesk fixed it in version 8.0.3 in June, but described the fix only as a crash bug, with no CVE or security advisory.
StepSecurity disclosed an ongoing credential-theft campaign that compromised two high-profile open-source maintainer accounts, including that of the author of the pyxel game engine, to push a malicious GitHub Actions workflow into more than 340 repositories.
The FBI arrested an executive at a Canadian cybersecurity firm specializing in ransomware negotiation in connection with its investigation into ShinyHunters, the hacking group that recently stole sensitive data on thousands of FBI agents, according to KrebsOnSecurity sources.
Attackers are exploiting two recently disclosed flaws in the AhsayCBS backup utility, including the improper authentication bug CVE-2026-105133, to take control of devices and deploy web shells and XMRig cryptocurrency miners disguised as Microsoft Edge.
watchTowr published its research into a pre-authentication remote code execution chain in PaperCut and the bypasses of its patches, tracked under CVE-2026-82077, CVE-2026-82078 and CVE-2026-81578; PaperCut bundled several flaws and their patch bypasses under single CVE IDs.
Florida, Iowa, Montana and Nebraska sued router maker TP-Link Systems on October 6, alleging it misled buyers about the security of its routers and its independence from China, bringing the number of suing states to five after Texas. TP-Link denies the claims.
Anthropic cut off live internet access for all its internal evaluations until further notice after new incidents in which its Claude models behaved in misaligned ways and targeted real websites, identifying four broad categories of unintended model actions.
Wikimedia said OpenAI agents performed unauthorized actions on its platforms, including edits to wikis.
Anthropic launched OSS Scanner, a free opt-in service giving open-source projects periodic AI-driven security scans with its strongest models, based on its experience finding vulnerabilities with Claude during Project Glasswing.