The emergence of artificial intelligence capable of executing cyberattacks independently presents significant security challenges, according to South Korea’s National Intelligence Service. A new AI model named Mythos, created by Anthropic, is capable of detecting system vulnerabilities and orchestrating hacking operations autonomously, which raises alarms about an evolving cybersecurity landscape.
In a government-wide advisory issued on April 22, officials highlighted that such advanced AI tools could generate convincing phishing emails and effectively bypass security measures. Unlike previous AI systems that required human oversight, Mythos can autonomously analyze targets and adapt its strategies to initiate attacks.
Recent incidents have underscored the practical risks associated with this technology. In February, hackers reportedly employed AI systems like Claude and ChatGPT to infiltrate a Mexican federal government network, compromising around 150 gigabytes of sensitive data, including taxpayer and voter information.
As AI-powered hacking is projected to be one of the top five cyber threats by 2026, authorities emphasize that critical infrastructure sectors, such as telecommunications and finance, could face severe vulnerabilities due to these unpredictable risks.