00:00Anthropic, but it's loosened its central safety policy to keep pace in a rapidly changing field.
00:05Now, this is after the Pentagon threatened to invoke a Cold War era law to compel Anthropic
00:10to allow the U.S. military to use the company's technology if it failed to comply with the
00:15government's terms, according to sources.
00:17Premier's senior tech editor, Mike Shepard, joins us.
00:20You go into the blog that Anthropic has pointed out, and yes, they talk about some of the
00:24steps that are already taken in terms of safety, some of the achievements they've already made,
00:27but they really highlight that despite rapid advancements in AI capabilities in the last
00:32three years, government action in AI safety has been slow.
00:36Is that what they're trying to call out here?
00:39You know, Carol, that is one thing they're trying to call out, but they are also acknowledging
00:43the stiff competition that they face in the AI space globally.
00:47They are competing with OpenAI, XAI, and of course with Google for a piece of what Martha
00:54Norton was also acknowledging as a massive competitive landscape.
00:58There is so much business to be done out there, and we have seen industry after industry shaken
01:03by the prospect of some of these new tools from Anthropic and from other providers that
01:08could really change the way business is done and how people actually do their work from
01:13day to day.
01:14So there is both that question, but then also the one that you zeroed in on, and that is
01:19that they are not seeing any traction at the federal level when it comes to discussion of
01:24these safety-oriented issues.
01:25And of course, that's the conversation coming up now at the Pentagon.
01:29The company is insisting that it will not be relaxing those two standards, that it doesn't
01:35want mass surveillance of Americans via the Pentagon through its technology or the fully
01:42autonomous use of its technology in weapons.
01:46But again, this relaxation of that safety principle when it comes to competition is a
01:53significant and worthy change to note here.
01:56Take us inside the room.
01:58We reported that Anthropic CEO Dario Amadei met with Secretary Hegsef, I believe, yesterday,
02:05Tuesday.
02:06There is a deadline of Friday from the Pentagon's perspective.
02:09But in our reporting, Anthropic has conditions of its own that it is putting to the Pentagon
02:14quickly.
02:16Yes, Anthropic has set those conditions as no mass surveillance and no use of its technology
02:22and fully autonomous targeting.
02:25And now those conditions, it's unclear at what point they will yield.
02:29The Pentagon says, look, you don't need those conditions.
02:31We will abide by the law.
02:33And we do not want, and this is according to the Pentagon's own new AI acceleration strategy
02:39released last month.
02:40They don't want companies to attach asterisks or conditions on the use of technology by
02:46the Pentagon if they are doing business with the government.
02:49And they have threatened the company with some dire consequences.
02:52One is that it would use the Defense Production Act to simply, in essence, impound the technology
02:57and use it over Anthropic's objections.
02:59Or failing that, they would actually declare Anthropic a supply chain risk.
03:05And that would mean that vendors with the Pentagon would have to certify that they are not using
03:09Anthropic's technology.
03:10And that would pose a business risk to Dario Amadei and his business.
Comments