00:00For the details, we're joined now by Bloomberg's Michael Shepard. So, Mike, I guess the idea here is that the
00:06administration is going to do something, yet this wouldn't necessarily go as far as they could have gone.
00:12Well, that's right. And that was some of the speculation this week that they may actually take that extra step
00:17of requiring the biggest AI developers to send their cutting edge models to a government agency for review and approval
00:25on safety and security grounds before their release.
00:28Now, that would be, of course, a little bit off brand for the Trump administration, which has really done everything
00:34it can to clear the regulatory underbrush for companies in a range of industries.
00:39Yet they are under some pressure to respond to concerns, especially around security related to artificial intelligence.
00:46Some of those were prompted in the past month by the release of Anthropik's new mythos model.
00:52So they have taken this step and they believe that by augmenting some of the preexisting information sharing programs that
01:00they have on cybersecurity to include AI companies,
01:03that they would get some of those concerns addressed and obviate the need, Kaylee, for having some sort of mandatory
01:11review program.
01:12Well, and that's what I wanted to ask, if this would essentially be formalizing something that already is to at
01:17least a certain extent underway in that.
01:19Don't we know that mythos, for example, is already being looked at by federal agencies inside the government already?
01:27Well, at least one federal agency and a key one, the National Security Agency, which, you know,
01:31is the tip of the spear when it comes to cyber defenses for the U.S. government, has been exploring
01:38its use.
01:38And we had that scoop recently that they were looking at what mythos could find about gaps in Microsoft products
01:47across the government.
01:48So there is some work on that front already.
01:51The White House is pushing for wider access to mythos for federal agencies.
01:56But so far, we're not getting any indication that that has come.
01:59And it's unclear whether this order would even address this directly, but it certainly could pave the way for that.
02:06And we do know that White House Chief of Staff Susie Wiles, who is taking a more direct role in
02:11crafting AI policy,
02:13including the memo that we're talking about right now, that she had met with Dari Amadei, the CEO of Anthropic,
02:21to talk about mythos together with some other administration officials.
02:26So the gears are turning.
02:27The White House has also, though, tried to cool the jets at Anthropic when it comes to releasing mythos to
02:34another group of trusted organizations,
02:37saying that they were concerned that letting it outside of the few dozen from Wall Street and Silicon Valley that
02:43have had a chance to review it
02:44and test it on their networks would be taking too much of a chance.
02:48And one thing that they're worried about, Kayleigh, is sharing it with partners who could then share it with foreign
02:55countries, companies, including some adversaries.
02:59All right. Bloomberg's Michael Shepard. Great stuff. Thank you so much.
Comments