13 hours ago
Objectivity seen as key to screening AI weapons
The Defense Ministry has compiled guidelines on ensuring appropriate human involvement in the research and development of defense equipment using artificial intelligence.
The guidelines are expected to cover R&D activities for equipment such as unmanned combat-support drones and ships, but ensuring objectivity and reliability remains a key challenge, since ministry officials are responsible for screening these activities.
The effectiveness of the guidelines also depends on how much AI data related to intellectual property private-sector companies disclose during R&D programs.
The United Nations is discussing how to regulate lethal autonomous weapon systems (LAWS), which use AI to identify and select targets without human involvement.
The Japanese government maintains that it does not intend to develop fully autonomous lethal weapons or conduct R&D on defense equipment banned under international or domestic law.
The guidelines include checks for compliance with international humanitarian law, clear assignment of human responsibility, prevention of overreliance on AI and fairness, meaning efforts to prevent discriminatory outcomes caused by uneven data use.
Screenings will be carried out by a panel made up mainly of officials from the Acquisition, Technology & Logistics Agency, an affiliate of the Defense Ministry, along with a council of experts within the ministry.
During the technology review process, the ministry will seek input on risk management from external experts familiar with advanced technologies, but it will make the final decisions on whether individual AI-related defense R&D projects are appropriate.
While the Foreign Ministry takes part in international discussions on rules for LAWS, it does not participate in the screening process. A Defense Ministry official noted, "We regularly hold discussions and share our views on LAWS with the Foreign Ministry."
As part of the screening process, private-sector companies involved in R&D may be asked to disclose the AI learning data and algorithms they use. Since the guidelines are not legally binding, one option could be to include such disclosure as a condition in the contracts.
"Through dialogue with the defense industry, we'll consider an appropriate way while striking a balance, including how to protect companies' intellectual property rights," a senior Defense Ministry official said.
