Japan Considers Advanced AI for Active Cyber Defense

Illustration. (Pexels)

Japan is considering using advanced artificial intelligence to strengthen its cyber defenses. AI is considered to face the threat of increasingly sophisticated attacks, including models that are able to find security gaps in a very short time.

Kyodo News, quoted on Monday, August 10, reported that the Director of Japan's National Cyber Yoichi Iida said that the use of advanced AI in cyber defense is increasingly difficult to avoid.

"It's not surprising that the time between the discovery of a security hole and its exploitation for an attack is only one-hundredth or one-thousandth of before," said Iida.

These concerns have emerged as AI models with high capabilities to find security vulnerabilities, including Claude Mythos from US startup company Anthropic.

After being introduced in April, Anthropic had limited access to the model to technology companies such as Google LLC and a number of financial institutions because it was concerned that its capabilities could be misused for cyberattacks.

Japan will start implementing a "cyber defense" policy on October 1 after Prime Minister Sanae Takaichi's cabinet made the decision in March.

The policy allows the government to take early action to prevent serious cyber attacks that could disrupt critical infrastructure, economic activities, and people's lives.

Iida said the government needed to consider the extent to which cyber defense operations could be automated. However, the final decision must still involve humans.

He has not explained the concrete form of the use of AI in the policy.

Iida also reminded Japan not to rely on one AI model alone.

He cited the decision of the United States government in June to stop access to the Mythos-class model through export control policies.

According to Iida, the termination of access has an impact on scanning security gaps in a number of important government systems.

The risk could be greater if cutting-edge AI technology is used in state-sponsored cyberattacks.

Iida said such attacks could become "much more sophisticated" and the intensity was expected to increase.