Threat Analysis: SCKit Worm Targets MSP ScreenConnect Instances, AI-Assisted Exploitation Goes Autonomous

Threats & Attacks

Threat Analysis: SCKit Worm Targets MSP ScreenConnect Instances, AI-Assisted Exploitation Goes Autonomous

This is your Daily Threat Analysis for the 16th of November 2026.

Two items today. One is an active campaign that is already inside the supply chain of hundreds of small businesses via their IT providers. The other is a research finding that changes the underlying economics of cyberattack. Together, they paint a picture of a threat environment that is moving faster than most UK SMBs are configured to respond to.

SCKit: The Worm Using Your MSP as the Entry Point

Between the 3rd and 13th of November 2026, threat actors enrolled more than 2,200 unauthorised guest sessions across 118 ScreenConnect servers belonging to 109 managed service providers and organisations. The detail that matters most: a significant proportion of connections originated from a single IONOS infrastructure provider, which gives you a sense of how coordinated this was.

The sessions do not announce themselves. They disguise themselves as legitimate computers using names drawn from cryptocurrency exchanges, KUCOIN-SRV, BACKUP-PROBIT and similar, and display screenshots of trading desktops to blend into the noise of a busy MSP environment. A technician scrolling through active sessions would have little obvious reason to flag them.

The underlying vulnerability is CVE-2026-84869, a flaw in cloud-hosted ConnectWise ScreenConnect instances. Once inside, the attackers deploy SCKit, a malicious worm that establishes persistence through scheduled tasks and spreads to customer-managed endpoints, meaning the businesses that the MSP supports, not just the MSP itself.

This is the supply chain problem in its most direct form. The small accountancy firm or the regional solicitor who outsources IT to a managed service provider has no direct relationship with ScreenConnect. They have never heard of CVE-2026-84869. And yet their endpoints are potentially compromised because their provider’s remote support tooling was vulnerable.

Microsoft published analysis of the SCKit campaign on the 13th of November. Huntress documented the session-hijacking component separately. The advisory attributes the initial access technique to abuse of ScreenConnect’s legitimate guest session functionality, not a complicated social engineering operation, not a nation-state zero-day. A known vulnerability in widely deployed software, exploited at scale.

What to do if you use an MSP:

Ask your IT provider, in writing, whether they use ConnectWise ScreenConnect. If they do, ask what steps they have taken to address CVE-2026-84869, when those steps were completed, and whether they have reviewed their ScreenConnect session logs for unauthorised guest connections dating back to the 3rd of November. If they cannot answer those questions promptly and clearly, escalate. If your IT provider tells you this does not affect you because you are too small, ask them how many of the businesses affected by this campaign thought the same thing before their endpoints were enrolled in a worm’s persistence mechanism.

If you are an MSP reading this: the Huntress and Microsoft advisories contain indicators of compromise. Review your session logs now.

AI Turns a Security Patch Into a Working Exploit, With No Human in the Loop

Google has confirmed that an AI agent was able to analyse a security patch and produce a working exploit from it without any human assistance. The Register reported this today.

I want to be precise about what this means and what it does not mean, because the coverage will range from dismissive to hysterical.

What it means: the gap between a vendor publishing a patch and an attacker having a functional exploit for the unpatched version has historically required human expertise, time, and effort. Reverse-engineering a patch to understand what it fixes and then constructing a working exploit from that understanding is skilled work. It could take days. Sometimes weeks. That gap, imperfect as it was, gave organisations time to patch before active exploitation began.

An AI agent that can compress or eliminate that gap changes the calculus entirely. It does not require the attacker to have reverse-engineering expertise on staff. It does not require significant time investment. The economics of exploit development shift in ways that make opportunistic exploitation of freshly patched vulnerabilities far more accessible.

What it does not mean: this is not, by itself, a fully autonomous attack chain. Generating an exploit is one step. Deploying it, maintaining access, achieving objectives, those remain more complex. But the direction of travel is clear, and the ScreenConnect story above is a useful illustration of where this leads in practice. The ARTEX framework, documented in an FBI advisory this month, is already being used to automate intrusions from initial reconnaissance through to data exfiltration. Some attacks documented in that advisory progressed from initial access to objective completion faster than many organisations run their weekly patch cycle.

The operational implication for UK SMBs is straightforward, if uncomfortable: the assumption that you have a week or two to apply a critical patch before active exploitation begins is no longer reliable. It may not have been reliable for some time. AI-assisted exploit development accelerates a trend, faster exploitation of known vulnerabilities, that was already well underway.

What this means in practice:

Patching is not optional and it is not a monthly exercise. Critical and high-severity patches, particularly for internet-facing systems, remote access tooling, and anything with a known vulnerability being actively discussed in security research, need to be applied within 24 to 72 hours. If your IT provider or MSP cannot demonstrate that capability, ask why not and what it would take to get there. If the answer is that it costs more, weigh that against the cost of an incident.

The Connection Between These Two Stories

SCKit is not a sophisticated nation-state campaign. It is opportunistic exploitation of a known vulnerability in widely used software, deployed at scale against a sector, managed service providers, that provides leveraged access to hundreds of downstream businesses per compromised instance.

The Google research on AI-assisted exploit development tells you why the window between vulnerability disclosure and active exploitation is narrowing. WardenStealer, documented separately this week, illustrates another dimension: attackers are now specifically targeting AI credentials, ChatGPT sessions, API keys, Copilot tokens, because those credentials have value in their own right and can be used to further automate subsequent attacks.

The pattern across this week’s intelligence is consistent. Attackers are reducing the manual effort required at every stage of the attack chain. They are targeting the chokepoints, MSPs, AI tools, remote access software, that give them the most leverage per compromise. UK SMBs sit downstream of all of these.

If Threat Analysis is useful to you, follow the show wherever you listen so tomorrow’s briefing lands automatically, and pass it to someone who needs the heads-up. The people who most need this information are rarely the ones who go looking for it.

Sources

SourceTitleLink
Microsoft Security BlogSCKit abuses ScreenConnect to deploy remote access and target technician sessionsLink
HuntressAttackers use Crypto Exchange Names to Disguise Unauthorized Sessions on 100+ ScreenConnect ServersLink
The RegisterGoogle: AI agent turned a security patch into a working exploit with no human helpLink
CyberintWardenStealer Malware: Turning AI Access into an Underground CommodityLink
BleepingComputerMicrosoft: Windows Server 2016 reaches end of support in 60 daysLink

Filed under

  • msp-security
  • supply-chain-risk
  • remote-access
  • smb-security
  • vendor-risk
  • nation-state-attacks
  • business-risk