“We Patched It” Is Not a Security Posture
Last week we covered AI agents running with root access, and the recurring theme was scope. Give an agent a narrow job and narrow permissions and it stays contained. Give it a broad job and broad permissions, and the outcome is whatever the agent decides is helpful that day.
This week the same idea shows up somewhere far less expected. In about ten days, one vendor shipped a security fix, watched it get exploited anyway, then shipped two more fixes that were broken on arrival. That is not a story about one company making mistakes. It is a story about what “we patched it” is worth as a security claim.
The Patch That Shipped With a Hole in It
N-able makes N-central, the remote monitoring and management platform that IT departments install so they can patch thousands of endpoints without visiting each one. On November 10, the company released build 2026.4.1.26 to fix a set of flaws researcher Sean Bales had disclosed. They were named as a group: Able2Shell.
Three days later, Bales published proof-of-concept exploits for two new issues, with names that were clearly a joke about how many times this would keep happening. 4EverAble2Shell gives unauthenticated remote code execution with root privileges on the N-central server. Agent2Shell chains together more than ten additional zero-days to run commands on managed Windows agents as NT AUTHORITY\SYSTEM.
The proof-of-concept worked against the build N-able had shipped on November 10. The patch for Able2Shell did not fix 4EverAble2Shell.
N-able’s advisory says the server issue depends on a non-default configuration in which the Legacy API is enabled and exposed to the public internet. Fine, probably true. But Ryan Dewhurst, founder of the threat intelligence firm Previdian, told The Hacker News that targeted N-central probing for 4EverAble2Shell was observed within hours of the proof-of-concept going up. There was no evidence of successful exploitation at the time of writing, which is the only good news in this paragraph.
The attack chain shows how a “we fixed it” story actually gets exploited. The threat actors went for a recently disclosed Apache Axis flaw, CVE-2026-90436 with a CVSS score of 9.8, first. That was the entry point. Then 4EverAble2Shell was the privilege escalation, and finally they pulled a shell script from that same IP address using curl. Dewhurst called it the first documented exploitation attempts against CVE-2026-90436, whose patches shipped November 6. Four days between the patch and the probes.
Patched, And Compromised Anyway
The ScreenConnect story is the more instructive one because the vendor’s patch actually worked.
Blackpoint Cyber’s Adversary Pursuit Group found that the SCKit malware campaign is not exploiting a new vulnerability. It is exploiting CVE-2026-84869, a CVSS 9.9 flaw in ConnectWise ScreenConnect that lets an attacker file-transfer and execute through an active remote session without host confirmation. The flaw is already patched, and Blackpoint confirmed that testing against a fully patched client blocks the payload. That is a good outcome, and it is still not sufficient.
Because the toolkit does not need the vulnerability. It hijacks authenticated operator sessions and uses legitimate ScreenConnect management features to push malicious code, running through the platform’s own Run Command functionality as NT AUTHORITY\SYSTEM. It includes a component the researchers named RelayTrap, which watches for the technician who shows up to investigate the suspicious endpoint and reverse-pivots into that technician’s workstation. Then it hunts the browser data and process memory for an authenticated session, the .ASPXAUTH cookies, so the attacker can reach other customer environments the same technician administers.
Blackpoint put it plainly: patching addresses the known exploit, but it does not eliminate the risk associated with a compromised operator session. The trusted path was the attack path.
The scale came from a second set of eyes on the same campaign. Huntress, which had visibility into its own customer base, counted at least 288 hosts across 48 organizations hit within 90 minutes, about a third of them in the first ten. Their telemetry points at a third-party ScreenConnect extension, RMM+, versions 1.0.31 through 1.0.32, whose authentication flaw let attackers run commands without valid credentials or MFA. That is a different door than CVE-2026-84869, opened in the same hour.

The Bug That Was Twenty-Two Years Old
Then there is the one that should make you rethink your relationship with version numbers.
PostgreSQL released fixes for 52 security flaws in November. The headline one is CVE-2026-6460, a type confusion flaw with a CVSS score of 8.8. A database user who can create objects can run code on the server, with the rights of the operating system user running PostgreSQL. Versions before 18.7, 17.12, 16.16, 15.20, and 14.25 are affected. Security researcher Mehmet Ince published a working exploit and deliberately omitted the final step, the one that makes the server run a command.
The root cause traces to May 5, 2004, when PostgreSQL gained the ability to change a column’s type. Twenty-two years. The flaw was not a new bug. It was a decision made in 2004 that nobody revisited, in a feature path that eventually became reachable by anyone who can create a table.
Ince adapted the exploit for managed PostgreSQL services and privately shared proofs of concept with their security teams in August. The providers he worked with include AWS (Aurora and RDS), Microsoft (Azure PostgreSQL), Google (AlloyDB), Supabase, Databricks, and Ubicloud. The writeup does not give each service’s patch status. If your database runs on a managed provider, the vendor patched the upstream project and you have no visibility into whether your instance got it. Someone else applied it.
The Framework That Already Exists
CISA’s Binding Operational Directive 26-04, “Prioritizing Security Updates Based on Risk,” issued June 10, 2026, replaces BOD 22-01 for Federal Civilian Executive Branch agencies. It requires those agencies to align vulnerability management against four criteria: Asset Exposure, Known Exploited Vulnerability Status, Exploit Automation, and Post-Exploitation Technical Impact. CISA has publicly encouraged all organizations to adopt it.
Notice what those criteria assume. Not that patch installation equals risk reduction, but what the vulnerability touches, whether attackers are using it, whether the exploit can be automated, and what they get if they succeed. Volume is why that matters. Microsoft shipped 1,259 fixes on November 10, its second consecutive month above a thousand, bringing the 2026 total past 5,100 CVEs according to Satnam Narang of Tenable. One of them is CVE-2026-124194, an Outlook flaw flagged as actively exploited and rated only Moderate 6.5. It connects to an attacker-controlled location without the usual external-content warning, disclosing an NTLMv2 hash that can be used to impersonate the user.
An email that leaks your hash, plus a stolen session cookie, plus a management platform running as SYSTEM, plus a trust path your technician’s browser hands over automatically. That is what the gap between “patched” and “hardened” buys you.
What To Do About It
- Verify patches by asset, not by ticket. Check the running version on the actual machine, not the deployment tool’s success report. A patch can ship and still leave the same class of bug reachable.
- Rank your queue against CISA’s four criteria, not CVSS severity. Asset exposure, known exploitation, exploit automation, post-exploitation impact. Start with what you actually have exposed to the internet, then what is already being exploited in the wild. Most breach paths start at a system you forgot was reachable.
- Strip legacy features from internet-facing systems. N-able’s Legacy API is off by default but was still exposed on probed servers. Audit version-specific toggles on management and monitoring platforms.
- Assume trusted operator paths can be stolen. For remote support and RMM tools, require phishing-resistant MFA on the operator side, scope sessions, and alert on session reuse from new locations.
- Treat local privilege and session material as crown jewels. The PostgreSQL exploit needed only the ability to create objects. The ScreenConnect campaign wanted browser cookies and process memory. Audit who can create database objects, and encrypt session storage on admin workstations.
The Real Measure
A patch ticket is a record of an action. It is not evidence of an outcome. The N-central probe wave started the same week as the fix, the ScreenConnect attackers kept going after the vulnerability was closed because they had moved to the session, and the PostgreSQL flaw had been reachable since 2004.
What separates a patched system from a hardened one is not the update you shipped. It is whether you verified it landed, whether you know what it was exposed to, and whether you closed the other doors attackers walked through instead. Those are three different jobs, and only the first one is a patch.

Sources
- Microsoft Patches 1,259 Flaws, Including Exploited Outlook Bug That May Expose NTLMv2 Hashes
- New N-central Exploits Work Against Patched Servers, Enable Root RCE and Windows SYSTEM Access
- SCKit Malware Turns ScreenConnect Support Sessions Into Potential MSP Attack Paths
- PostgreSQL Patches 52 Flaws, Including 22-Year-Old Bug Enabling Authenticated RCE
- SCKit Worm Installs Rogue ScreenConnect Clients and Session Cookie Relays Across Hundreds of Endpoints (Huntress, Nov 11, 2026)
- SCKit: The Worm That Turns MSP Trust Into a Supply-Chain Weapon (Aviatrix Threat Research Center)
- CISA Binding Operational Directive 26-04: Prioritizing Security Updates Based on Risk
- CISA Known Exploited Vulnerabilities Catalog
Seiller Secure IT
Want the specific answer for your own network?
Every article here is written for owners and IT leads who need the plain version of a security problem. If you want the specific version for your environment, a security assessment maps what you actually run, what is actually exploitable, and what to fix first. It is a few days of focused work, not a quarter of consulting.
See the security assessment service
Prefer to talk it through first? Use the contact form on the homepage: https://seillersecureit.com/#contact