The National Intelligence Service. Photo by Asia Today
April 22 (Asia Today) — South Korea’s intelligence agency has warned that artificial intelligence capable of independently carrying out cyberattacks is emerging as a potential “game changer,” raising concerns over a new era of cyber threats.
The National Intelligence Service said it has issued a government-wide security advisory after identifying growing risks tied to advanced AI models that can autonomously detect vulnerabilities and execute hacking operations.
According to the agency, a next-generation AI model known as “Mythos,” developed by U.S. artificial intelligence company Anthropic, demonstrates capabilities that go beyond traditional tools. The system can independently identify system weaknesses, design intrusion pathways and generate malicious code in real time.
Officials warned that such tools could craft highly convincing phishing emails and bypass security systems, making them viable for real-world cyberattacks.
“What we are seeing is no longer AI assisting hackers, but AI acting as the hacker itself,” the agency said in its advisory distributed to government bodies on Sunday.
The key difference, officials said, lies in autonomy. Earlier AI systems functioned as assistants, helping write code under human direction. In contrast, Mythos can analyze targets, adapt strategies and execute attacks without continuous human input.
The agency cited reports that the model uncovered a long-undetected vulnerability in OpenBSD – a security-focused operating system – and even designed a potential exploitation route.
Authorities say the threat is no longer theoretical. In February, unidentified hackers reportedly used AI models such as Claude and ChatGPT to breach a Mexican federal government system, exposing approximately 150 gigabytes of sensitive data. The leaked information included taxpayer records, voter data and government employee credentials.
Investigators believe AI tools were used to locate vulnerabilities, automate attack scripts and streamline data extraction.
The intelligence agency had already identified “AI-powered hacking” as one of the top five cyber threats for 2026, describing it as a shift in the cybersecurity paradigm. Officials warned that AI’s involvement across the entire hacking process could create unpredictable and difficult-to-control risks.
Critical infrastructure – including telecommunications, energy and financial systems – could become primary targets, with potential consequences extending beyond data breaches to large-scale system disruptions, the agency said.
The government is considering talks with Anthropic later this month to discuss countermeasures and response strategies. As of now, no confirmed cases of domestic institutions or companies being directly affected by the Mythos model have been reported.
An agency official said that while AI has so far been a supporting tool for hackers, misuse of security-focused models could pose serious risks to public services and national infrastructure.
“Thorough preparedness is essential,” the official said, adding that authorities are still working to fully assess the capabilities of emerging AI cybersecurity models.
— Reported by Asia Today; translated by UPI
© Asia Today. Unauthorized reproduction or redistribution prohibited.
Original Korean report: https://www.asiatoday.co.kr/kn/view.php?key=20260421010006726
Click Here For The Original Source.
