deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

AI exploit tools and Microsoft's record 1,252-CVE Patch Tuesday fuel a patch panic

Microsoft's November update fixed over 1,000 CVEs by vendor counts while AI attack frameworks compressed exploit timelines to days or less, and security watchers warn that blanket panic patching is becoming its own risk.

AI exploit tools and Microsoft's record 1,252-CVE Patch Tuesday fuel a patch panic

IT teams were told to "patch everything, right now" this month, after Microsoft shipped one of the largest security updates in its history in the same week that three widely deployed web technologies disclosed serious flaws. A dev.to analysis argues that the resulting panic is now a problem in itself, and lays out which fixes actually deserve priority.

AI tools are collapsing the patch window

According to the dev.to write-up, an FBI FLASH warning issued on November 13 states that criminals have used the open-source AI framework ARTEX since at least August 2026, and that some intrusions reached admin-level access in under three hours. ARTEX has been linked to the theft of personal data on roughly 67,000 people at South Korean financial institutions, and to a ConnectWise Cloud Backup compromise that produced sign-in attempts against 2,250 customer tenants.

Google's Mandiant, cited in the same piece, groups ARTEX with Hexstrike and CAIRN under the label agentic offensive frameworks, and reports the first case it has seen of attackers using a framework, called exploit_agent, that compares a patch against the old code and drafts an exploit with no human in the loop. Cloudflare says more than 50 families of AI pentesting tools scanned its customers within three days, with such traffic up 30-fold, and Mandiant's M-Trends 2026 report puts average time-to-exploit at minus seven days, meaning attacks often start before a fix even exists.

A Patch Tuesday like no other

Microsoft's update landed on November 10. Qualys counts 1,252 CVEs fixed; ManageEngine counts 1,094, with 63 rated Critical. The tallies differ, but both dwarf November 2025's total of 63. Microsoft averaged about 74 CVEs a month from November 2025 to March 2026, and roughly 650 a month since April 2026, with the last three releases each fixing close to 1,000 or more.

Much of that growth is attributed to AI finding old bugs in old code. A chart by Patrick Garrity shows 229 patches released against 7,349 findings on Anthropic's Frontier Red Team 90-day track as of November 16, and Anthropic is cutting its disclosure window from 90 to 30 days for high and critical bugs, starting with reports received from January 1, 2027. The flood is expected to grow, not shrink.

The bugs to fix first

Five CVEs stand out among the thousand-plus fixes:

  • CVE-2026-124194: an Outlook flaw that leaks NTLMv2 password hashes when a crafted message file is opened. Already exploited and on CISA's KEV list, with U.S. agencies expected to patch by November 24. The fix ships through Office's own update channel, for Microsoft 365 Apps and Office LTSC 2021 and 2024, not Windows Update, so verify it installed. Anyone who may have leaked a hash should reset their password, admins first, then finance and identity teams.
  • CVE-2026-119958: unauthenticated LDAP remote code execution on domain controllers, CVSS 9.8.
  • CVE-2026-116928: a Windows Search flaw triggered by an attachment Exchange indexes, CVSS 9.8, requiring both the Windows and Exchange updates.
  • CVE-2026-118207: Exchange Server remote code execution, with exploitation rated more likely.
  • CVE-2026-118146: Active Directory Domain Services remote code execution, also rated more likely.

The month also ends fixes for .NET 8, .NET 9, PowerShell 7.4 LTS and 7.5, and brings the final security update for Windows 11 23H2 Enterprise and Education. Dev.to recommends planning moves to .NET 10 and PowerShell 7.6.

Three hacks in three days

Microsoft was not the only trigger. On November 16, an attacker abused a flaw in Optimizely's account signup, which let brand-new accounts act as administrators of existing customers, to inject malicious JavaScript into the Project JavaScript served to visitors of 178 organizations, including Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock. The payload swapped cryptocurrency wallet addresses, skimmed payment cards, stole cookies and screenshots, and loaded the Coruna and DarkSword iOS exploit kits in a hidden iframe aimed at older iPhones. It was rebuilt 26 times during the day and netted roughly $7,040 by 19:25 UTC. Optimizely blocked delivery at 16:06, closed the signup flaw at 16:10 and removed the injected code by 22:24 UTC. Cards typed into hosted payment iframes like Stripe and Adyen were out of reach, and CSP policies mostly failed because many sites allow unsafe-eval. Optimizely Web users are advised to check their project code and rotate cookies.

On November 17, Java4Shell hit Java's built-in Base64 decoder, affecting Java 17 and newer on Intel and AMD servers; update Java or disable the fast decoder. On November 18, NGINX2Root affected nginx with HTTP/2 enabled; update to 1.30.7 or 1.31.9, or turn HTTP/2 off.

Why it matters

Resilient Cyber has coined the term Patchmageddon for patching everything AI finds without checking what matters. GreyNoise data shows fewer than one in 100 exploit attempts against headline bugs actually succeed, so blanket patching spends effort where risk is low while the genuinely urgent items, like verifying the Outlook fix installed and resetting exposed credentials, can slip. With attack timelines at or below zero and published patches now doubling as exploit roadmaps, prioritization and verification, not raw speed, are what keep networks safe.

  • #security
  • #microsoft
  • #patch-management
  • #ai
  • #vulnerabilities

Related posts