The group behind the Aur0ra ransomware has been using AI agents to move through victim networks, collect credentials and work its way up to domain administrator access, according to Sophos.
AI-assisted intrusion workflow (Source: Sophos)
The company, which tracks the group as GOLD OGRE, recovered artifacts from the Cursor AI coding assistant in late August that show the attackers handing the agents VPN logins, proxy access, targets and objectives. The agents then mapped the network and Active Directory, hunted for passwords in network shares, documents and backups, and moved laterally with little input from the people directing them.
In some networks, the agents gained domain management access, and three of those victims were later named on Aur0ra’s data leak site. Breaking in and deploying the ransomware stayed under human control.
Prompts written in Russian
The researchers also obtained some of the operators’ instructions to the agents. In one, an operator said the Windows AD network needed testing and offered to set up a SOCKS5 proxy and user accounts for the agent.
Another instructed it, “Don’t try brute-forcing domain administrator accounts anymore. Try other, less noisy, methods and vectors.” A third asked it to find the backup servers, NAS servers and virtualization systems in the victim’s infrastructure.
Operator logs also contained keyboard slips typed in Cyrillic, which the company says suggest the operators are Russian speakers.
A red team cover story
To get around the guardrails built into the AI tools, the attackers described the intrusions as authorized red team activity or diagnostic testing. The attacks ran through Cursor’s Composer agentic coding model and models from Anthropic, xAI and Moonshot.
“Analysis indicates that the threat actors initially used Anthropic almost exclusively but then introduced other models. The shift may be related to the threat actors’ objectives, as they transitioned from building tools to conducting intrusions.,” Sophos noted.
The company’s Counter Threat Unit assesses with high confidence that GOLD OGRE used AI to test whether its tools could evade EDR products, then gave associates similar technology for Aur0ra attacks. The researchers consider it highly likely that other ransomware groups will build AI agents into their operations too.
Gambit Security and CloudSEK published their own analyses of Aur0ra operations, both apparently based on the same exposed open directory. Gambit’s report mentions a Claude 4.5-Sonnet agent being deployed in semi-automated attacks.
“AI can accelerate phases of the attack chain and reduce the level of technical aptitude needed to conduct an attack. Existing controls such as patching internet-facing devices, deploying phishing-resistant multi-factor authentication (MFA), and monitoring EDR alerts are still effective, but successful defense relies on a quick response,” Sophos concluded.
——————————————————–
Click Here For The Original Source.
