Hackers Used AI Agents to Break Into Thousands of ConnectWise Customers

A breach of ConnectWise, one of the largest vendors of remote IT management software, allowed attackers to spy on thousands of companies that use it. Microsoft attributes the intrusion to a China-aligned actor that relied on AI agents to move through ConnectWise's customer base at speed.

ConnectWise's software is the kind that managed service providers install on every machine they look after, precisely so they can reach into those machines remotely. Compromising it does not get an attacker into one company. It gets them a legitimate, privileged channel into everyone that company's IT provider serves.

The automation is the story

Supply-chain attacks through IT management vendors are not new. What is new is the throughput.

Historically the bottleneck in an intrusion of this type was human: once inside the vendor, an attacker still had to work through downstream environments one at a time, each with its own layout, deciding manually what was worth taking. That labour cost imposed a natural limit, and it bought defenders time.

Agents remove the limit. The same access now yields thousands of simultaneous, individually-tailored intrusions, which is why the reported victim count is in the thousands rather than the dozens.

Where this lands in Addis Abeba

Ethiopian firms rarely buy ConnectWise directly. They do, increasingly, outsource IT to managed service providers, and those providers run exactly this class of tooling. The exposure in a supply-chain breach is inherited, not chosen — which means the security question a finance director should be asking is not about their own network but about their vendor's.

It arrives in a week that has made the same point from several directions: Interpol recovered 105 million dollars in a cybercrime operation across 21 African countries, arresting 1,462 suspects and exposing losses above 385 million dollars; the developer platform Railway told all its users to revoke and rotate private keys after attackers decrypted some customer secrets; and four major US banks reported unauthorised AI agents reaching customer accounts.

Open

Microsoft has not published the victim list, and ConnectWise has not said how the initial access was obtained or how long the attackers were resident before detection. Nor is it clear which agent frameworks were used, or whether existing detection tooling registered the activity as anomalous at all.

Sources: TechCrunch, 22 October 2026, TechCrunch, 22 October 2026, Birr Metrics, 23 October 2026