00:00Next, the U.S. Defence Secretary gave Anthropic an ultimatum this week.
00:04Open its artificial intelligence technology for unrestricted military use by Friday
00:09or risk losing its government contract.
00:12Officials also warning that they could designate the firm as a supply chain risk
00:17or invoke a Cold War-era law called the Defence Protection Act.
00:22For more, let's cross live now to France 24's Wassim Kournay, who's standing by for us in L.A.
00:29Wassim, tell us first, what exactly is the Pentagon demanding here from Anthropic?
00:37Yeah, the Pentagon is essentially requesting that Anthropic remove its restrictions
00:42from how its AI model, which is called CLOD, can be used inside a new internal network
00:50either do that or face the consequences.
00:53The Defence Secretary, Pete Hexeth, has given the company a Friday deadline tomorrow at 5 p.m. Eastern time
00:59If Anthropic doesn't agree, the Pentagon says it is ready to sever ties with the company
01:04to terminate its contract and even label it a potential supply chain risk,
01:09which would effectively put Anthropic on a government blacklist.
01:13Officials have also raised the possibility of invoking something called the Defence Production Act,
01:18which we'll get to in just a moment.
01:20Right now, Anthropic places limits on how its system, CLOD, can be used.
01:25Those limits restrict things such as mass surveillance or autonomous weapons.
01:30It is the last major AI firm not to be fully plugged into this new Defence Department system.
01:35The CEO of the company, Dario Amadei, has publicly raised concerns about unchecked military use of AI.
01:42Now, the Pentagon, of course, has said it has no plans to break the law,
01:45but also that it will not allow a private company to set operational limits.
01:51Now, in the last few minutes, we've heard back from Anthropic.
01:53We were waiting to hear back from them following that new request by the Defence Department.
01:57They said that they cannot in good conscience exceed to the Department of Defence's request.
02:03It also explains that decision, saying that language received from the Department of Defence made no progress,
02:10that in terms of the mass surveillance of Americans and the use of autonomous weapons,
02:17saying that the words, this new language issued by the Department of Defence,
02:21is surrounded by legalese, which would allow those safeguards to be disregarded.
02:26And, Waseem, as you say, the Defence Production Act has been mentioned as a possible means of retaliation.
02:31Just how significant would it be if the Pentagon actually were to use it?
02:39Yeah, it would be a massive escalation.
02:41Let's talk a bit about what the Defence Production Act is.
02:44It dates back to the 1950s, to the Korean War,
02:47and it allows the federal government to direct private industry to prioritize national defence needs.
02:54It's been used in the past for things like COVID, medical supplies, energy shortages,
03:00but it's never been used to force a private company to alter safety rules of a product that it designed
03:07itself.
03:08Legal experts say that's the difference, and it could be a major issue if the Defence Department does decide to
03:13go down that route,
03:15that the government, it does allow it to change, excuse me, that it can force a company to prioritize needs,
03:22but it cannot force a company to change AI rules or functions,
03:28and that that would immediately be met, likely, with court challenges.
03:32And there is also tension within the Pentagon as to how to approach this,
03:36because, as I did mention, the government is threatening Anthropic to cut ties with it
03:43and to put it on a supply chain risk list.
03:45But the officials, privately within the Pentagon, have also acknowledged that it is critical,
03:51the company Anthropic is critical to its military operations.
03:54Wasim, thanks so much for joining us with that update and with that breaking news on that story.
03:59That's Wasim Kournei joining us live there from L.A.
Comments