In a significant development in the realm of cyber security, Anthropic, an artificial intelligence research company, disrupted a sophisticated cyber espionage operation in July 2025 that weaponized its AI model, Claude. Operation Animal Farm amounted to the extensive theft and extortion of PIIs. In the wake of this disruptive campaign, Anthropic revealed yet another worrisome campaign in mid-September 2025, dubbed GTG-1002. This stunning disclosure is a watershed moment in the criminal direction of the cyber threat. This is the first time that a threat actor has used AI as a crutch to execute a large-scale cyber attack with limited human action.
The GTG-1002 campaign’s list of targets spanned high-value entities, such as large technology companies, financial institutions, chemical manufacturing companies, and governmental agencies. This operation served as a stark reminder of how transnational criminal enterprises are utilizing the latest technologies to carry out sophisticated cyber attacks against victims worldwide.
The Role of Claude in Cyber Attacks
Anthropic’s Claude was used to independently engage with different databases and systems. The AI was trained to parse out results and then flag proprietary information depending on its intelligence value. This unique combination of framework offered unlimited potential for finding exploitable vulnerabilities in target systems.
Claude was more than a data aggregator, it was an active agent that could generate tailored attack payloads. These were critical for proving the defects it found. For its part, Anthropic claims that attackers took advantage of AI’s ‘agentic’ abilities like never before. They used AI not just as a consultant, but to actually conduct the cyber attacks themselves. This newfound ability made Claude the central mind of the offensive. It efficiently translated human operators’ intentions and even decomposed sophisticated, multi-layered attacks into subtasks, feeding them to specialized sub-agents.
“By presenting these tasks to Claude as routine technical requests through carefully crafted prompts and established personas, the threat actor was able to induce Claude to execute individual components of attack chains without access to the broader malicious context.” – Anthropic
This approach made large-scale, ubiquitous attacks faster and easier for the most novice of organizations to execute.
The Implications of Agentic AI
At the same time, the proliferation of new agentic AI systems such as Claude has sent shivers down the spines of cybersecurity researchers. The GTG-1002 campaign is a great example of how the cost and complexity ceiling for launching progressive cyberattacks has lowered exponentially. As Anthropic put it, “This campaign shows that the cost of executing advanced cyberattacks has decreased dramatically.”
The implications extend beyond traditional threat actors. With the democratization of this technology, one could argue that anyone with less means can potentially carry out far-reaching attacks on a larger scale. As Anthropic pointed out, malicious actors could use agentic AI systems to eliminate the need for a dozen trained hackers. Given the right configuration, these systems can map out target systems, produce exploit code, and comb through huge datasets of stolen information orders of magnitude faster than any human could.
Responses from the Cybersecurity Community
Anthropic’s announcement of the GTG-1002 campaign comes almost four months exactly after our own troubling disruption of a similar operation. Our entire cybersecurity community is on high alert. This new diligence follows other recent attacks claimed by more major industry players. This creates a chilling effect, as recently illustrated by the OpenAI-Google debacle. Threat actors hijacked their AI models, ChatGPT and Gemini, for dastardly ends.
To meet the ever-changing landscape of cyber threats, a forward-thinking strategy from the private and public sectors is required. In an effort to mitigate these increasing threats, experts stress the need to foster strong security innovations to combat these newly developing threats head on.