Japan eyes cutting-edge AI under pre-emptive cyber defence policy

Japan is mulling the use of advanced AI systems as part of its preemptive cyber defence strategy, according to Kyodo News. The technology is seen as essential to counter the increasing threat of sophisticated cyberattacks, particularly from models like Anthropic's Claude Mythos. This AI, known for its capability to identify security weaknesses, initially had restricted access due to fears of potential misuse.

Japan's National Cyber director, Yoichi Iida, highlighted the rapid acceleration in the timeline between discovering vulnerabilities and their exploitation by these advanced AI models. The government's move to permit active cyber defence measures from October aims to preemptively eliminate serious cyber threats that could disrupt critical infrastructure, economic activities, and people's livelihoods.

Iida emphasized the importance of balancing automation in cyber defence operations with human oversight in final decision-making. He also cautioned against over-reliance on a single AI model, citing the US government's recent export control directive that suspended access to model classes similar to Claude Mythos. Such restrictions could potentially intensify the sophistication and intensity of nation-sponsored cyberattacks, should any occur.