Artificial Intelligence & Machine Learning,
Next-Generation Technologies & Secure Development
Google Identifies Iranian and Chinese Threat Groups as Most Active Users of AI Tools
Recent disclosures from Google reveal that Iranian and Chinese threat actors are leveraging the company’s artificial intelligence application, Gemini, for reconnaissance and vulnerability discovery. Attempts by some groups to circumvent the application’s built-in security measures have also been noted.
According to Google’s statement, advanced persistent threat (APT) groups from over 20 countries have utilized Gemini in the past two years to gather intelligence on payload creation, malicious scripting, and exploit evasion techniques, with the highest levels of activity attributed to actors from Iran and China.
More than ten Iranian threat groups have accessed Gemini, with APT42 being highlighted as particularly active. This group has reportedly solicited information on crafting phishing emails and conducted reconnaissance on defense and policy targets. APT42, also identified as Mint Sandstorm or Cobalt Illusion, has ties to the Islamic Revolutionary Guard Corps.
APT42’s inquiries extend to vulnerabilities in various technologies, including Mikrotik routers, Apereo identity solutions, and Atlassian products, demonstrating their research into exploitation methods and server-side request forgery. The group has been noted for seeking guidance on employing generative AI tools for offensive operations, specifically in training red teams on AI’s applications in cyberattacks.
Meanwhile, Chinese state-sponsored activities have primarily focused on reconnaissance efforts, evidenced by one group’s unsuccessful attempts to reverse engineer components of the Carbon Black endpoint detection and response server. Another entity has sought assistance in signing Microsoft Outlook plugins for covert deployment, while APT41 has examined Gemini’s infrastructure in their explorations.
The activities of APT41, tracked as Brass Typhoon and Earth Baku, tie into a broader campaign linked to the ongoing compromise of Sophos firewalls, underscoring the persistent risk of edge device exploitation for espionage operations. Other users of Gemini include North Korean and Russian hacking groups, with North Korean actors primarily investigating job-related operations on LinkedIn to sustain their IT workforce scams. Russian activities appear to be geared towards enhancing domestic technological capabilities and developing AI chatbots leveraging large language models.
In addition to their malicious activities, threat actors have attempted to bypass Gemini’s security guardrails to exploit Google products and create harmful scripts. Google’s use of publicly available prompts mitigated many of these malicious attempts, illustrating the challenges of securing advanced AI applications against state-sponsored misuse.
