Ransomware · AI Security
3AM ransomware encrypts 37 UK school systems in suspected AI-assisted attack
A ransomware group encrypted 37 systems at a UK school estate, with researchers suspecting the attackers used AI tools to automate parts of the attack. The incident highlights new risks as criminals experiment with AI.
Ross Grant2 min read
The short version
A group of criminals broke into a UK school's computer systems and locked up 37 of their machines using ransomware. Security experts think the attackers may have used AI tools to help automate some steps, but it's not clear exactly how much AI was involved. The attackers stole passwords, made secret accounts, and deleted backups to make recovery harder. This attack shows that criminals are starting to experiment with AI in real attacks.
By the numbers
- 37school systems encrypted by ransomwareBleepingComputer, Hackread
- 47total systems in the school estate targetedHackread
- 3Active Directory domains breached and linkedBleepingComputer, Hackread
What happened?
A ransomware group used 3AM ransomware to encrypt 37 systems at a UK school estate, according to Bitdefender and Hackread. Researchers suspect the attackers used AI tools to automate parts of the attack, but the exact role of AI is unconfirmed. The attack took place between 10 and 19 September 2026. The attackers stole credentials, created backdoor accounts, and deleted backups before deploying the ransomware.
How the attack worked
A 3AM ransomware affiliate targeted a UK school estate and managed to encrypt about 37 out of 47 systems, according to Bitdefender (opens in a new tab) and Hackread (opens in a new tab). The attackers gained access through an unmanaged host and used a mix of standard Windows tools and public hacking utilities. They extracted and decrypted credentials from backup servers, copied browser password databases, and used tools like Mimikatz to steal more credentials. The attackers then mapped out domain relationships, searched for references to other schools, and created two backdoor administrator accounts named 'library' and 'svc_bkpmon.'
To expand their reach, they set up two-way trusts between three previously separate Active Directory domains. This allowed them to move laterally and access more systems. In the final stage, they tried to remove endpoint protection software and deleted system-state backups before deploying the 3AM ransomware, which encrypted the majority of reachable machines.
Suspected role of AI
Bitdefender researchers noticed that many commands were repeated with slight changes at regular intervals of 40 to 100 seconds. This pattern suggested the use of an "agentic loop," where an AI tool (possibly a large language model) could be generating and adjusting commands based on feedback. However, the researchers could not confirm whether the process was fully automated or if a human was copying and pasting between an AI tool and the compromised systems. The evidence points to a hybrid operation, with both human and AI involvement.
This is not the first time 3AM ransomware has been linked to advanced tactics. The group has previously used email bombing and fake IT support calls to breach networks. The suspected use of AI in this attack highlights a growing trend where attackers experiment with automation to speed up or improve their intrusions.
The response
Bitdefender recommends that organisations separate privileged accounts across domains, monitor for new domain trusts, protect backup infrastructure, and ensure every machine has up-to-date endpoint protection. The UK's National Cyber Security Centre (NCSC) has also warned that AI-powered ransomware attacks are likely to become more common by 2028, urging critical infrastructure to prepare.
What you should do
Schools and other organisations should review their domain trust relationships, monitor for unusual account creation, and secure backup systems. Regularly updating and testing endpoint protection can help block or limit the impact of ransomware. Monitoring for repeated or automated command activity may also help detect early signs of AI-assisted attacks.
Claimed vs confirmed
- Unconfirmed
Attackers used AI tools to automate parts of the ransomware attack.
Bitdefender researchers observed behaviour suggesting possible AI assistance, but could not confirm the exact role or type of AI used.
Why it matters
This incident shows that ransomware groups are beginning to use AI tools to automate parts of their attacks. Even if the AI involvement was limited, defenders should prepare for more automated and complex attacks in the future.
Key points
- 37 systems at a UK school estate were encrypted by 3AM ransomware.
- Researchers suspect attackers used AI tools to automate command retries.
- Attackers stole credentials, created backdoor admin accounts, and deleted backups.
- Bitdefender recommends separating privileged accounts and monitoring domain trusts.
- AI-assisted attacks are expected to become more common in coming years.
Sources
BleepingComputer
Reference
Hackread
Reference
