US AI Firm Anthropic Says It Stopped Massive Cyberattack Led by Chinese Hackers Using AI Tools
Anthropic claims it disrupted a first-of-its-kind AI-driven cyberattack targeting global organisations.
In a startling revelation, US-based artificial intelligence company Anthropic, the developer of the Claude chatbot, announced that it successfully prevented what it called the first large-scale AI-driven cyberattack executed with minimal human involvement. The company alleged that Chinese state-sponsored hackers manipulated its AI systems to target around 30 organisations across multiple sectors worldwide.
In a detailed statement, Anthropic described the operation as a “highly sophisticated espionage campaign” that unfolded in mid-September 2025. The attackers allegedly used Claude’s autonomous features to identify vulnerabilities, write exploit code, and infiltrate systems across industries such as finance, chemical manufacturing, technology, and government agencies. According to the firm, the operation demonstrated how advanced chatbots can function as semi-autonomous agents in cyber warfare.
Investigations revealed that the hackers camouflaged their activity by posing as a legitimate cybersecurity firm conducting defensive tests. They reportedly used “jailbreaking” methods to bypass built-in safeguards, allowing Claude to perform restricted tasks like scanning network infrastructures and extracting sensitive information. The company found that the AI executed nearly 80 to 90 percent of the attack sequence independently, with minimal oversight from human operators.
Also Read: Chinese Woman Floods Hotel Room Over Dispute, Forced to Pay 280 Times the Price
Once the activity was detected, Anthropic acted swiftly, initiating a 10-day internal investigation to trace the breach, alert compromised entities, and work with international cybersecurity agencies. It blocked affected accounts and collected intelligence to prevent similar incidents in the future. The company noted that although some operations failed due to AI misjudgments—such as generating inaccurate credentials—the incident demonstrated the growing capability of autonomous AI systems in both defensive and offensive contexts.
In its findings, Anthropic warned that this event signals a new era in cybersecurity, where malicious actors could exploit AI to perform complex attacks that once required teams of skilled hackers. The company urged global coordination to develop stronger safeguards and frameworks governing AI “agents,” stressing that the barrier to executing sophisticated cyber campaigns has now dropped significantly.
Also Read: Chinese Moms Spend $100K a Year on US Schools Despite Trump’s Tough Policies
