US and Russia Remove Requirement to Verify AI-Identified Targets From Draft UN Rules

The United States and Russia removed from a draft UN document on autonomous weapons requirements that systems operate “predictably” and “reliably,” as well as a mandatory human review of AI-identified targets before a strike.

The Washington Post reported this, citing its own sources and documents it obtained.

In early September, hundreds of diplomats gathered in a large UN conference room in Switzerland to work on an agreement that was intended to regulate the use of lethal autonomous weapons and become the first international agreement addressing AI-powered combat systems.

On the final day of negotiations, US and Russian diplomats removed several provisions from the document aimed at ensuring the safe use of artificial intelligence in weapons.

According to sources familiar with the discussions, the text no longer included requirements that systems operate “predictably” and “reliably,” or that ethical considerations be taken into account when using AI-enabled weapons.

The US and Russian delegations also removed a provision requiring human verification of AI-identified targets before a strike.

During a closed-door session, after UN cameras were switched off and observers from civil society organizations were asked to leave, the US and Russia each brought in about ten lawyers. That was nearly twice the number of lawyers on other delegations. According to three sources, the changes were being made so quickly that smaller delegations were unable to keep up with the process.

The negotiations are taking place amid an intense public debate over the threats AI development could pose to humanity’s existence. Leaders of technology companies are calling on world leaders to jointly establish limits on the technology. The UN has been trying for several years to regulate the use of AI-enabled weapons but so far has made only limited progress.

Meanwhile, efforts to introduce rules in the United States have stalled, with the Trump administration arguing that any regulation of AI would stifle innovation.

“The United States also categorically rejects any attempts to create a globalist scheme to control artificial intelligence, which is being discussed so much at present – it will henceforth officially be called ‘superintelligence,’” US President Donald Trump said.

As previously reported, during the active phase of military operations against Iran, the US came close to attacking a Chinese vessel that allegedly was carrying components for a nuclear weapons program. The incident reportedly stemmed from an erroneous intelligence report prepared with the help of AI.

Only shortly before the planned operation did officials more carefully review the report prepared by an analyst with the US Special Operations Command. They discovered that, in response to the analyst’s query, a chatbot had combined open-source intelligence with classified signals intelligence available to the government and reached an incorrect conclusion about the vessel’s cargo. The analyst then used AI again to format the findings as a standard intelligence report and circulated it.

According to one of CNN’s sources, the report was “completely false.” At the same time, the source emphasized that it “nearly started a war.” Any US operation against a Chinese vessel could have escalated into an armed conflict between the two countries.