Cybersecurity researchers uncover an AI driven intrusion ecosystem used by ransomware affiliates to target dozens of organizations worldwide.
According to findings from SOCRadar reported by The Hacker News, cybersecurity investigators have uncovered an artificial intelligence driven intrusion ecosystem operated by affiliates of the INC and The Gentlemen ransomware groups. The operation relies on an autonomous red team kill chain framework named Phantom, managed by a threat actor assessed to be Russian speaking. This threat actor acts as both an intrusion and data extortion operator while providing infrastructure support to different partners across the wider ransomware ecosystem.
The campaign centers around an AI executed and human approved operating structure. A human operator provides a victim specific prompt and reviews every step of the process, while the intrusions themselves are executed through OpenCode with OpenRouter hosted commercial models. An examination of an exposed directory at 45.227.254.43 revealed that the activity logged 8,257 messages across 133 sessions. When models refused to act on specific prompts, operators were found rewriting requests as authorized red team assessments rather than switching models.
Initial access is achieved by exploiting public facing edge and remote access systems at scale. SOCRadar noted that attackers targeted FortiClient EMS gateways, Ivanti, Citrix, SonicWall, and OpenVPN gateways, SimpleHelp remote management servers, and cPanel hosting controls. Once inside, the operators followed a recurring lifecycle of enumerating the domain, escalating to domain admin privileges, harvesting credentials with stealer tools, and prioritizing sensitive files from finance, HR, legal, and executive departments over reverse-SOCKS tunnels.
Credential reuse served as a major force multiplier during the attacks. In one specific instance, a single reused password unlocked a victim organization's VPN, servers, remote monitoring and management tools, backups, and security management platforms all at once. The threat actors deployed a diverse toolkit that included bespoke frameworks, modified open-source tools, compiled binaries, a Cloudflare fronted Mythic agent named QuickLine, and a Quick Heal or Seqrite EDR killer abusing the Zemana driver for a bring your own vulnerable driver attack.
The operation adopted a data centric approach to extortion, examining stolen financials to determine the maximum amount a victim could pay without involving banks or insurers. Attackers then sent masked proof samples, ramped up pressure through scripted vishing calls, and threatened to notify customers, partners, and regulators. Stolen data was published on data leak sites if victims failed to meet a 24 to 72 hour deadline. SOCRadar estimated that 62 organizations were targeted by the campaign, with the United States accounting for more than 50 percent of the targets, primarily impacting industrial, manufacturing, energy, healthcare, and pharmaceutical sectors.
Background
The disclosures regarding the Phantom framework and associated ransomware affiliates stem from exposed infrastructure analyzed by threat intelligence analysts at SOCRadar. The investigation detailed how the Russian speaking operators utilized AI managed knowledge bases to track victims, log compromised credentials, and monitor campaign progress across 133 sessions and thousands of logged messages. While the final financial outcomes of the ransom demands remain unknown due to unrecovered cryptocurrency wallets, four of the 62 targeted organizations were ultimately listed on data leak sites operated by INC and The Gentlemen.
Why It Matters
The discovery of the Phantom framework illustrates a concerning evolution in how threat actors leverage artificial intelligence tools for cybercrime. By integrating commercial AI models into an automated kill chain while keeping a human operator in the loop for approvals and prompt adjustments, attackers can scale their intrusion operations with unprecedented efficiency. This blurring of lines between automated intelligence and manual oversight presents significant challenges for traditional enterprise defense systems, particularly when attackers exploit widespread edge device vulnerabilities and simple credential reuse to bypass multi-layered security controls.
Illustration: AI-generated image, not a photograph of the event.