Google and Anthropic Warn of Rising AI-Powered Cyberattacks

Google and Anthropic are warning that artificial intelligence is rapidly changing the cybersecurity landscape, with malicious actors increasingly using AI systems to support sophisticated cyber operations. Recent findings indicate that AI can help attackers carry out complex activities at greater speed and scale, raising concerns among technology companies and security researchers about how quickly defensive systems must adapt.

Anthropic’s latest threat intelligence report identified several campaigns involving state-linked groups and cybercriminal organisations attempting to misuse its Claude models. The company said some operations involved espionage, phishing, malware-related activity and other forms of cyber abuse, while it worked to detect and disrupt the activity.

The warning comes as AI agents become increasingly capable of handling multiple tasks with limited human intervention. Security experts say this creates a new challenge because AI systems can potentially become both tools used in attacks and targets themselves, making traditional approaches to identifying and attributing cyber incidents more difficult.

Anthropic has also reported attempts to misuse its technology for activities extending beyond conventional cybercrime, including surveillance, influence operations and potentially dangerous biological research. The company says newer AI models require stronger safeguards because their capabilities can make harmful requests more consequential.

The developments are adding pressure on the technology industry and governments to strengthen AI security standards. More than 100 organisations, including major AI companies, have previously warned that the window for preparing against AI-enabled cyberattacks may be narrowing. As AI becomes more deeply integrated into business and government systems, experts say security measures will need to evolve alongside the technology.