I recently discovered the blog “The Stack”, when they quoted from one of my posts on CVE. In the short time I’ve been reading their posts, I’ve been very impressed by the breadth and depth of research they perform; they find important stories I’ve seen nowhere else. However, they outdid themselves with this story that appeared in today’s post. The story starts with these three paragraphs:
The US Department of War plans to connect Elon Musk’s Grok AI models – and others – to “every classified and unclassified network throughout our department,” according to Secretary of War Pete Hegseth.
Hegseth has also ordered all military departments to “update federated data catalogs exposing their system interfaces, data assets, and access mechanisms across all classification levels… [and] deliver their current catalogs” to the department’s Chief Digital and AI Office within 30 days.*
“Our data advantage is meaningless if our developers and operators cannot exploit it,” he said, launching a “transformative AI acceleration strategy” that aims to get AI agents making "kill chain" decisions and put AI models in the hands of Pentagon staff "at all classification levels."
I’ll let you read the rest of the article yourself, although this is a very good summary.
Let’s be clear: This is the same Grok that spews antisemitic content and undresses women and children. It seems that, after those wonderful “achievements”, Grok has received a well-deserved promotion. Now, it will have full access to all information (including classified information) on all Defense (War) networks. But don’t worry…Grok won’t misuse any of that information. Scout’s honor.
Moreover, “…DoW plans to build out AI agent capabilities to ‘unleash… Al enabled battle management and decision support, from campaign planning to kill chain execution.’” Clearly, the Department is going to encourage “warfighters” (formerly known as “soldiers”, “sailors” or “airmen”) to trust Grok to make decisions on whom to attack and how.
I recently wrote a post that distinguished between trusting AI as an extremely powerful intelligence source and trusting AI’s judgment. If a user understands this difference and makes sure that all consequential decisions remain firmly under human control, AI has an important role to play in control of the power grid and “warfighting”. However, it seems DoW has 100% trust in AI’s judgment, as well as its intelligence.
Near the end of that post, I illustrated the possible consequences of trusting the judgment of any automated system (whether or not it carries the “AI” label) by describing how, on September 26, 1983, Stanislav Petrov, a Lieutenant Colonel of the Soviet army, disobeyed his explicit orders and decided not to relay to his superiors the fact that the screen in front of him displayed one word: “Launch!” He did this not because he had solid information that the system behind the screen didn’t have all available information, because it did. He made his decision because of a feeling that this order was based on faulty information (which turned out to be the case).
Had Lt. Col. Petrov not disobeyed orders, there would almost certainly have been a nuclear war within 15 minutes. It would have left most of both the Soviet Union and the US destroyed and their populations mostly dead or dying, with toxic radiation clouds killing millions of people in other countries.
The current leadership of DoW prides itself on its certainty and its determination to move forward against all opposition. It’s highly unlikely that, if Lt. Col. Petrov were on duty in the US today and a similar situation were to come up, he would have the courage not to take a step he knows will lead to Armageddon.
Tom Alrich’s Blog, too is a reader-supported publication. You can view new posts for one month after they come out by becoming a free subscriber. You can also access my 1300 existing posts dating back to 2013, as well as support my work, by becoming a paid subscriber for $30 for one year (and if you feel so inclined, you can donate more than that!).
If you would like to comment on what you have read here, I would love to hear from you. Please comment below or email me at [email protected].
Tom Alrich’s Blog, too is a reader-supported publication. You can view new posts for one month after they come out by becoming a free subscriber. You can also access my 1300 existing posts dating back to 2013, as well as support my work, by becoming a paid subscriber for $30 for one year (and if you feel so inclined, you can donate more than that!).