Ransomware Operators Deploy Hybrid AI Model in UK School Attack
A 3AM ransomware affiliate attacked a UK school using conventional techniques—stolen credentials, domain trust manipulation, lateral movement, and data theft—but displayed signs of AI involvement. Bitdefender researchers identified behavioral patterns suggesting an agentic loop handled repetitive command generation and troubleshooting while a human operator made strategic decisions. Evidence included unusual Veeam backup API calls deviating from vendor documentation, numbered script revisions, and timing patterns showing machine-speed retries interspersed with human-paced task sequencing.
The attack represents a novel hybrid model—neither fully AI-generated, traditionally AI-assisted, nor autonomously AI-operated. Researchers lacked direct access to the attacker’s system to confirm the AI component, but behavioral signatures (repetitive probes, self-labeled outputs, consistent retry loops) suggest coordination between automated systems and human judgment. For financially motivated threat actors, this hybrid approach appears more practical than full autonomy, using AI to automate tedious troubleshooting while operators handle strategic decisions.
The attack relied on standard ransomware playbook techniques and didn’t require capabilities unique to AI. Agentic loops generate substantial noise through repetitive commands and brute-force probes, leaving detectable traces despite operational speed advantages. Standard ransomware defenses remain effective: isolate credentials across domains, protect backups, restrict remote execution, limit shell access, and deploy comprehensive endpoint security. Controls that block an action the first time work equally well against the hundredth.
Read the full article
Continue reading at Dark Reading →This is an AI-generated summary. Read the original for the full story.