The Department of Defense has already banned artificial intelligence from launching nuclear weapons autonomously. A bipartisan group has taken extra precautions to ensure that AI cannot launch nuclear weapons.
As was announced earlier in the week, Senator Edward Markey, Representatives Ted Lieu, Don Beyer, and Ken Buck, have introduced the Stop Nuclear Launch by Autonomous AI Act. This act would codify the existing Pentagon rules regarding nuclear weapons. These rules, effective as of 2022 read:
By the same token the bill states that no autonomous system, without meaningful human supervision, can launch a nuke weapon or “select targets or engage them” with the intent of launching a nuke. The text states that “any decision to launch nuclear weapons should not be taken by artificial intelligence.”
Why introduce this bill if it is already prohibited? The bill’s sponsors point out that the 2021 National Security Commission on Artificial Intelligence report recommends affirming a prohibition on autonomous nuclear weapon launches. This is not only to stop it happening within the US government, but also to encourage similar commitments from China or Russia. The bill’s public release will draw attention to the dangers of the current generation of autonomous artificial intelligence systems. This is a concern that is shared by both and the tech industry. As indicated in the press release it is also an opportunity to showcase the sponsors’ efforts to prevent nuclear proliferation, such as a recent bill that restricts the president’s power to declare war unilaterally. Let’s let you tell the obvious War Games prank.