AI cyberattacks set to outpace human hackers, experts warn

Cybersecurity experts are starting to warn that AI-powered attacks are escalating their threat because Anthropic has developed its Mythos model, which can find security weaknesses at a speed that exceeds human hackers.

The threat affects governments, corporations, and critical infrastructure worldwide, raising concerns about how to defend against AI-driven exploits in real time.

According to a leaked blog post, AI developer Anthropic has warned that its upcoming Mythos model will operate automatic software vulnerability scanning and exploitation at exceptional speed. OpenAI has issued a similar warning in December about future models posing high cybersecurity risks.

Cato Networks Founder and CEO Shlomo Kramer commented on this: “The agentic attackers are coming. This is a watershed event in the history of cybersecurity.” According to experts, a single AI agent can beat hundreds of human hackers with its continuous scanning, testing, and exploitation of vulnerabilities.

A Russian-speaking hacker used multiple AI tools like Anthropic’s Claude and China’s DeepSeek to hack over 600 devices in 55 countries in January. According to Amazon Web Services’ security research team, this was done using AI tools. According to Gambit Security Director of Threat Intelligence Eyal Sela, AI made complex hacks easy for attackers, giving them "superpowers" regardless of their skills.

Claude was also reportedly used in attacks on Mexican government agencies in February, where sensitive voter and tax data were leaked.

AI can also be used to improve cybersecurity. Twenty Co-founder and CEO Joe Lin commented: “Attackers need only one way in, while defenders must cover every surface. AI enables faster monitoring and automated patching, but humans must control the outcomes.”

Experts highlight how AI-based cybersecurity tools will be essential in combating the development of autonomous AI attacks.

Anthropic and other AI development centres are conducting tests on their models with select organisations in preparation for large-scale AI-based cyber threats.