Korea to launch sovereign AI this year: Science minister

Korea will launch a sovereign AI model specializing in cybersecurity by the year's end to respond to growing security threats in the digital space, the science minister said Thursday.

Minister Bae Kyung-hoon's comments come amid growing concerns over digital security after Washington abruptly imposed export controls on its advanced AI models, including Anthropic's Mythos 5, which specializes in vulnerability detection.

"We are pushing to create an AI model specializing in cybersecurity within this year by training our existing sovereign AI model on security-related data," Bae said during a policy briefing held at the Blue House, presided over by President Lee Jae Myung.

The minister noted that the current level of Korea's sovereign AI is not enough to respond to evolving cybersecurity threats helped by generative AI, reaffirming previous calls that the country should review developing a frontier model, on par with Mythos, in the long run.

"In my humble opinion, Korea could build one, too," he said.

But the number of GPUs the ministry can supply to companies making efforts to develop advanced foundation models is still insufficient, Bae argued, as he requested further budgetary support from the government.

During the briefing, officials also discussed ways to institutionalize "white hat hacking," or legally breaking into computer systems under the owner's consent to identify vulnerabilities.

The science ministry is making preparations for legislation that would provide legal grounds to perform ethical hacking on companies without their consent for vulnerability detection under certain conditions, a Science Ministry official said.

Korea was ranked third in terms of AI competitiveness from an international ranking agency, Bae said. He expects the country could be elevated to second place in the next round of assessment to be announced in August, he added.

Yonhap