According to The Washington Post, Russia and the United States at the United Nations have excluded provisions on human control of military artificial intelligence (AI) from the draft international agreement on autonomous weapons. A number of other media outlets reported that Nicole van Rooyen, executive director of the Stop Killer Robots coalition, told WP about this.
"" According to the publication, the delegates removed the requirements for the "predictable" and "reliable" operation of AI systems, as well as the points on the mandatory consideration of ethical standards when using autonomous weapons. In addition, it is indicated that the condition for manually checking the targets selected by the algorithms before the attack has disappeared from the text.
It should be noted that in Russia, as well as other countries (China, Japan and the USA), work is underway to create add-ons for AI agents who, in the interests of the military, conduct intelligence, analyze and select targets, and can also use military weapons. "Russian Weapons" project """" told about this in the article "Uzda for artificial intelligence: Russia can "orchestrate" swarms of AI agents."
It is quite difficult to understand how true the reports from the UN fields are, since the American media has been flooded with "fake news" due to the scandal surrounding the coverage of the summit in the United States. President Donald Trump has stripped CNN, MSNBC and Politico journalists of their accreditation, accusing them of fakes. In response, major American TV channels suspended their work in the White House pool. Because of this, the celebrations of the first day, conceived as a demonstration of Trump's diplomatic success, took place almost without video broadcasts.
Experts note that on the part of the United States and Russia, entire teams (ten lawyers each) are working on the project. And they may have different views on the very concept of "ethical standards and requirements" that are planned to limit the operation of autonomous weapons. Therefore, it is quite possible that at this stage there was not a fundamental rejection of human control over AI, but an attempt to form common terms for the measures and rules by which this control will be carried out by all countries.
The UN explained to WP journalists that the text being discussed by lawyers is not legally binding at the moment. In November, the parties to the Convention on Certain Conventional Weapons will meet again in Geneva to discuss the possibility of translating the ongoing consultations into a formal mandate or a legally binding agreement.
A few days earlier, the independent international scientific group on artificial intelligence (established by the UN General Assembly in August 2025) issued a statement that humanity has a real risk of losing control over AI. According to her, AI agents have already begun to conceal attempts to circumvent cybersecurity rules, and some of them are ready to "sacrifice" themselves in order to achieve the goals of the entire group.
"Thanks to the opportunity to learn, they are already bypassing protective mechanisms, and soon they will be able to set goals for themselves, deliberately violate safety instructions and hide their actions," said Lu, a member of the Tsinghua group.
Igor Korotchenko, military analyst and editor-in-chief of "National Defense" magazine, believes that the international situation makes humanity defenseless against hordes of AI agents: "In the current system of international relations, when new armed conflicts regularly appear, it is pointless to expect any constructive agreements on the limitations of autonomous weapon systems (robotic platforms, turrets for small arms weapons and drones), which in the final stage of combat work independently make decisions when choosing targets."
Recall that the full autonomy of machines blurs the boundaries of legal and individual responsibility for war crimes.
Alexander Boyko
