The military applications of Artificial Intelligence (AI) have already introduced great risks and challenges to the world. As such, we should be vigilant about lowering the threshold of war due to the development of military AI , and actively work to prevent avoidable disasters. The "Defense Artificial Intelligence and Arms Control Network" advocates that the design, research, development, use, and deployment of military AI throughout its whole life cycle should comply with the following principles:
- Military AI should be applied to national defense, and its applications in offense operations to other countries and regions should be opposed.
- Defense AI should adhere to the principle of minimum and necessity.
- Deployers of such technologies should strive to minimize unnecessary harm.
- We must forbid the use of lethal autonomous weapons (LAWS).
- We must ensure meaningful human control for all AI-enabled weapon systems, and avoid cognitive overload during Human-AI interactions.
- We must ensure the safety and security of military AI applications.
- We must ensure AI systems’ decisions are transparent and explainable to humans.
- We must ensure that military AI applications are trustworthy and reliable.
- Military AI systems should take measures to mitigate/avoid potential biases in data, AI algorithms and human operators, as well as avoid targeting specific groups of people, and ensure justice.
- Military applications of AI should be based on the principle of improving "humanitarianism", not undermining or challenging this principle. We must advocate the use of AI in military operations for "humanitarian relief", without harming other lives.
- Strict technical and ethical assessments of military AI systems should be conducted in accordance with the principle of responsibility, and effective regulations and responsibility tracing mechanisms should be established.
- We must prevent the proliferation of AI weapons.
- Military AI systems should be in full compliance with international and humanitarian law.
- AI-enabled military systems and domains (e.g., chemical weapons, nuclear weapons, etc.) should still fully comply with the norms of their respective fields.
- Relevant personnel using military AI systems should ensure adequate training, correctly grasp how to use the systems, and accurately understand the consequences of using the systems. Public awareness on the use of military AI systems and their risks should be continuously improved.
The above principles will be continuously updated with the consensus of the experts of "Defense AI and Arms Control Network". To join our expert group, or offer comments and suggestions, please email us at: email@example.com