Prime Minister Sanae Takaichi meets with Peter Thiel, chairman of Palantir Technologies Inc. (Asahi Shimbun file photo)

EDITORIAL: Can autonomous decisions be made using foreign AI tech?

（社説）自衛隊とＡＩ　主体的判断を守れるか

The Defense Ministry has decided to introduce artificial intelligence in the Self-Defense Forces’ command and control.

Such a move may have been unavoidable given the rapid spread of AI utilization in military affairs around the world.

But discussions must continue on such topics as how to ensure human involvement, what effect depending on a foreign computer system will have on autonomy and what form democratic control should take.

The U.S. military uses Palantir Technologies Inc.’s “Maven Smart System.”

That system rapidly integrates and analyzes a huge amount of data, such as satellite and drone images as well as communication information, selects targets and proposes the most effective attack method.

There are reports that the system was used in the military action in Venezuela as well as the attack on Iran.

It has also been pointed out that Palantir technology has been used by Ukraine in its counterattack against Russia.

The government insists that AI will only support analysis and operational planning and that final decisions will be made by humans.

But because it will be difficult for humans to reassess in a short time the recommendation made by AI, they may end up simply going along with that judgment.

AI is not infallible so there are doubts as to whether it can accurately differentiate between civilians and miliary personnel.

There is also the issue of hallucination in which AI provides a mistaken response in a plausible manner.

There is also the possibility of not being able to explain why AI made a mistaken judgment, leading to a “black box problem.”

Because Palantir’s AI has already been used in joint exercises involving the SDF and the U.S. military, it can be considered a leading candidate should the SDF decide to use such technology.

But if the computer system at the core of command and control is the product of a foreign company, albeit one from an ally, there is the slim possibility of suspension of service or functional restrictions being made by the company or the government of another nation.

That would affect national defense, which is a core component of state sovereignty.

Concerns have also been raised about information leakage because confidential data will be provided.

There is also the question of whether Japan’s legal structure based on an exclusively defensive posture can coexist with a U.S.-made product that places priority on military rationality.

At the same time, there is no telling if Japan can catch up with the technology and data accumulated by a U.S. company based on actual experience.

There are strong calls from within the government and ruling coalition for a domestic computer system so consideration has been given to using a combined system or to replace the system in the future with a domestic one.

There will be a need to demarcate what can be left up to a foreign nation and where an uncrossable line exists.

Introducing AI into command and control is not simply about using new technology, but is also an issue about state sovereignty and the character of its defense.

Hasty decisions should be avoided and thorough discussions held on such topics as the extent to which the technology will be used, where responsibility would lie, how humans would be involved and the level of dependence on a foreign nation.

--The Asahi Shimbun, July 15

防衛省が自衛隊の指揮統制に人工知能（ＡＩ）を導入する方針だ。ＡＩの活用は世界の軍事分野で急速に進んでおり、対応は避けられない。ただ、人間の関与をどう担保するのか。外国のシステムに依存した場合、自律性が損なわれないか。民主的統制のあり方を含めた議論が不可欠だ。
　米軍は、米データ解析企業パランティア・テクノロジーズの「メイブン・スマート・システム」を採用している。
　衛星画像や無人機映像、通信情報など、膨大なデータを素早く統合・分析し、攻撃目標の選定、最適な攻撃手段の提案などを行う。ベネズエラへの軍事行動やイラン攻撃に使われたとの報道もある。パランティア社の技術は、ウクライナのロシア軍への反撃などにも使われていると指摘される。
　政府は、ＡＩが行うのは分析や作戦立案への支援であり、最終判断は人間が行うとしている。ただ、人間がＡＩの判断を短時間で検証し直すことなど困難で、結局、追認することにならないか。
　ＡＩは完全無欠ではない。戦闘員と民間人を確実に識別できるのか。もっともらしく誤った回答をするハルシネーションもある。判断ミスが起きた際、なぜそうなったのかを説明できない「ブラックボックス化」の問題もある。
　パランティア社のＡＩは、すでに自衛隊と米軍の共同演習で使われており、自衛隊が導入する場合の有力候補のひとつではあろう。
　ただ、指揮統制の基盤となるシステムを、同盟国とはいえ、外国の企業に依存することになれば、他国の企業や政府の判断でサービス停止や機能制限が行われる可能性はゼロとはいえない。国防という国家主権の中核が左右されかねない。機密データを提供することによる、情報漏洩（ろうえい）への懸念も指摘される。軍事的合理性を優先する米国製が、専守防衛に基づく日本の法制と両立するかも問われる。
　一方で、実戦経験も踏まえ、米国企業が蓄積する技術やデータに日本が追いつけるかは見通せない。政府・与党内では、国産システムを推す声も根強く、組み合わせて使うことや、将来的に国産に置き換える案も検討されているという。どこまでを外国に頼り、どこが譲れない一線なのか。見極めが欠かせない。
　指揮統制へのＡＩ導入は、単なる新技術の採用ではなく、防衛のあり方や国家主権にかかわる問題である。拙速は避け、利用範囲や責任の所在、人間の関与のあり方、外国依存の是非などについて、徹底した検討が求められる。

(2026年7月15日)