Latest / Trumponomics / Will Mythos Ruin or Save the Global Financial System?

Transcript

Will Mythos Ruin or Save the Global Financial System? · Trumponomics

• 0:02
  Speaker 1: Bloomberg Audio Studios, podcasts, radio news.
• 0:08
  Speaker 2: As this tech evolves, we may need to have a
• 0:10
  Speaker 2: more standardized process for if there is a systemically important
• 0:14
  Speaker 2: development in the tech, there is an actual process for
• 0:17
  Speaker 2: deciding who gets access. It's done in a more democratic way,
• 0:20
  Speaker 2: in a more equal way, and that's a hard thing
• 0:22
  Speaker 2: to happen.
• 0:32
  Speaker 3: I'm Stephanie Flanders, head of Government and Economics at Bloomberg,
• 0:35
  Speaker 3: and this is trump Anomics, the podcast that, as ever
• 0:39
  Speaker 3: looks at the economic world of Donald Trump, how he's
• 0:41
  Speaker 3: already shaking the global economy and what on earth is
• 0:44
  Speaker 3: going to happen next. And this week's episode is being
• 0:47
  Speaker 3: recorded a bit early on Thursday, the sixteenth of April,
• 0:50
  Speaker 3: in Washington, in the middle of the Spring meetings of
• 0:53
  Speaker 3: the International Monetary Fund and World Bank, where an AI
• 0:56
  Speaker 3: model that few of the officials here had probably even
• 0:59
  Speaker 3: heard of a week ago is threatening to upstage even
• 1:03
  Speaker 3: the conflict in Iran as the subject everybody wants to discuss.
• 1:07
  Speaker 3: It all started with a Bloomberg story a week earlier
• 1:10
  Speaker 3: revealing that the US Treasury Secretary Scott Bessant had gathered
• 1:13
  Speaker 3: Wall Street bank leaders to warn about mythos the latest
• 1:18
  Speaker 3: AI model from Anthropic that it was so powerful that
• 1:21
  Speaker 3: it could herald a new era of cataclysmic cyber attacks
• 1:25
  Speaker 3: carried out autonomously by bot. That meeting called at short
• 1:30
  Speaker 3: notice set off alarm bells with central bankers and regulators
• 1:34
  Speaker 3: around the world, who all then rushed to arrange their
• 1:36
  Speaker 3: own meetings to understand how this model and others like
• 1:39
  Speaker 3: it could raise a new level of systemic threat. But,
• 1:43
  Speaker 3: as with most things AI, the answer turns out to
• 1:46
  Speaker 3: be a bit complicated because the same people who tell
• 1:48
  Speaker 3: you how dangerous this new model is, a model Anthropic
• 1:52
  Speaker 3: itself has said is so powerful it's only making it
• 1:55
  Speaker 3: available to a select group of firms, will also tell
• 1:58
  Speaker 3: you it's going to be the best security against cyber
• 2:02
  Speaker 3: attacks we've ever had. So frightened, relieved, perplexed, How should
• 2:07
  Speaker 3: we feel about mythos and does the stability of the
• 2:10
  Speaker 3: world financial system and the broader economy really rest on
• 2:13
  Speaker 3: the answer? While we thought it was worth getting some
• 2:16
  Speaker 3: kind of debate going on that with Laura noonan global
• 2:19
  Speaker 3: finance and regulation reporter who's been in DC chasing this
• 2:22
  Speaker 3: story but is normally based in London. Laura, welcome to
• 2:26
  Speaker 3: the show, Thank you, and Michael Dung is the geoeconomics
• 2:30
  Speaker 3: technology analyst at Bloomberg Economics, and he's previously served as
• 2:33
  Speaker 3: a policy advisor and presidential management fellow in the Chips
• 2:36
  Speaker 3: Program Office at the Department of Commerce.
• 2:39
  Speaker 4: Michael, welcome back. Great to be back.
• 2:44
  Speaker 3: Our colleagues actually wrote a big explanatory piece on the
• 2:47
  Speaker 3: day that we're recording, which sort of painted a picture
• 2:50
  Speaker 3: of the testers inside Anthropic, who over the last I
• 2:54
  Speaker 3: guess a few months have been waking up to the
• 2:57
  Speaker 3: extraordinary and potentially quite frightening power of this new model Mythos. Michael,
• 3:02
  Speaker 3: what is it? I guess we should get to basics
• 3:04
  Speaker 3: on what's special about.
• 3:05
  Speaker 1: It Mythos's Anthropics newest model, and in particular, the capability
• 3:09
  Speaker 1: that's set off alarm bells is its ability to autonomously
• 3:12
  Speaker 1: discover vulnerabilities on a scale that we haven't seen before.
• 3:16
  Speaker 1: So this capability in itself isn't brand new. If you
• 3:19
  Speaker 1: look at cloud Opis four point six, which is the
• 3:21
  Speaker 1: model immediately preceding this, they did show early signs of
• 3:24
  Speaker 1: this capability in terms of finding five hundred zero day exploits,
• 3:28
  Speaker 1: which are exploits that the developers didn't know about and
• 3:30
  Speaker 1: vulnerabilities that attackers could exploit.
• 3:32
  Speaker 3: So zero day is you have zero days to fix
• 3:34
  Speaker 3: them because you didn't know about them exactly.
• 3:36
  Speaker 1: Yes, and those tend to be the most devastating used
• 3:39
  Speaker 1: by hacker groups and even nation states, and generally the
• 3:42
  Speaker 1: ideas that many of them hold a repository of them
• 3:44
  Speaker 1: to use at certain times. Opus four to six already
• 3:47
  Speaker 1: demonstrated ability to find some of these. The difference with
• 3:49
  Speaker 1: Mythos is its ability to find many more of them
• 3:51
  Speaker 1: autonomously at speed we hadn't seen before. And in many cases,
• 3:55
  Speaker 1: one of the anecdotes was they had researchers who did
• 3:57
  Speaker 1: not have security training essentially instructed to run overnight and
• 4:01
  Speaker 1: the next morning when they came back, it had found
• 4:03
  Speaker 1: a lot of vulnerabilities. And it's that ability to find
• 4:05
  Speaker 1: vulnerabilities in much of the code and infrastructure that underpins
• 4:09
  Speaker 1: the Internet, our financial system, across every major operating system,
• 4:13
  Speaker 1: internet browsers, et cetera, that I think really is scaring people.
• 4:16
  Speaker 3: We're going to get into the double edged nature of this,
• 4:18
  Speaker 3: because if you can find the vulnerabilities, then maybe as
• 4:22
  Speaker 3: long as the system is in the right hands, you
• 4:24
  Speaker 3: could potentially fix them ahead of others. The reporting we
• 4:27
  Speaker 3: had on this, one of the paragraphs jumped out at
• 4:29
  Speaker 3: me when they were doing this testing. Is it Mythos
• 4:31
  Speaker 3: orchestrated the digital equivalent of a bank robbery, getting past
• 4:35
  Speaker 3: security protocolsm through the front door of networks, breaking into
• 4:38
  Speaker 3: digital vaults that gave it access to online treasures. AI
• 4:42
  Speaker 3: had picked locks, but now it could pull off an
• 4:44
  Speaker 3: entire heist.
• 4:46
  Speaker 4: Yes.
• 4:47
  Speaker 1: I think that's a great demonstration of where we are
• 4:49
  Speaker 1: in the frontier right And so previously, when you just
• 4:52
  Speaker 1: had the models alone, they were sophisticated. They could program,
• 4:54
  Speaker 1: they could write code well, et cetera. If you've heard
• 4:56
  Speaker 1: all the chatter over the past year Agantic systems, this
• 4:59
  Speaker 1: program that we've seen specifically in Mythos is a perfect
• 5:01
  Speaker 1: demonstration of that. The ability to act autonomously, chained together
• 5:05
  Speaker 1: independent actions. And I think even more so than finding
• 5:08
  Speaker 1: specific vulnerabilities, that ability to chain together minor exploits into
• 5:11
  Speaker 1: a significant attack campaign is what has researchers most concerned,
• 5:15
  Speaker 1: because that's the ability that's most difficult to guard against.
• 5:18
  Speaker 1: One of the most popularly circulated graphics, which was at
• 5:21
  Speaker 1: the top of Anthropics release article talking about Mythos was
• 5:25
  Speaker 1: looking at Firefox exploits in which they compare something like
• 5:28
  Speaker 1: Mythos found one hundred and eighty one ways to exploit
• 5:32
  Speaker 1: Firefox versus two to four point six. But the key
• 5:35
  Speaker 1: there was the one hundred and eighty one In many
• 5:36
  Speaker 1: cases relied on two vulnerabilities specifically, but Mythos specifically using
• 5:40
  Speaker 1: those two was able to find a variety of ways
• 5:42
  Speaker 1: to attack the browser, and that sort of demonstrates the
• 5:44
  Speaker 1: greater latitude of action that this model has versus predecessors.
• 5:48
  Speaker 3: In some ways, this is an unusual thing for us
• 5:50
  Speaker 3: to be talking about because we're getting into kind of
• 5:52
  Speaker 3: high tech, but of course we're also ai is infiltrating
• 5:55
  Speaker 3: so many different parts of our world. Now, that's why
• 5:57
  Speaker 3: we would have to talk about it. But Laura, when
• 6:00
  Speaker 3: you came to these meetings, you say in a piece
• 6:03
  Speaker 3: that you've been writing that you might have expected potential
• 6:07
  Speaker 3: for World War three, or certainly major developments in Iram
• 6:10
  Speaker 3: which are causing enormous economic ructions, to be the dominant
• 6:13
  Speaker 3: topic this week in Washington, But they have been a
• 6:16
  Speaker 3: little bit hijacked by Mythos.
• 6:18
  Speaker 2: Yeah, I mean in terms of the people who are
• 6:20
  Speaker 2: here from a pure financial sector perspectives. So we're talking
• 6:23
  Speaker 2: about the governors, the financial regulators, the banks themselves, the
• 6:27
  Speaker 2: people who are interested in the pure financial sector stuff.
• 6:29
  Speaker 2: The main risk everyone's been talking about has been mythos
• 6:33
  Speaker 2: and what it could mean for the financial sector, and
• 6:36
  Speaker 2: it has been fascinating to watch it evolve because these
• 6:38
  Speaker 2: are people who feel very well versed in understanding financial risks.
• 6:41
  Speaker 2: They could talk to you about the model underpinning bank
• 6:43
  Speaker 2: capital all day long. These kind of models are brand
• 6:47
  Speaker 2: new territory for them. I think they'll actually benefit a
• 6:49
  Speaker 2: lot from listening to our colleagues comments here, because the
• 6:52
  Speaker 2: level of understanding is just not there, and they read
• 6:54
  Speaker 2: this thing, they think it's big, it's scary. They don't
• 6:58
  Speaker 2: know how it's going to regain. And if you hear
• 6:59
  Speaker 2: the idea that it could do an entire bank heist
• 7:02
  Speaker 2: in like five minutes, ten minutes, that is terrifying for
• 7:05
  Speaker 2: the people whose job it is to protect the financial system.
• 7:08
  Speaker 3: Cyber risk cybersecurity must be something that you have been
• 7:11
  Speaker 3: hearing about in these circles for a while. I'm certainly
• 7:13
  Speaker 3: very struck. When you talk to the heads of banks
• 7:18
  Speaker 3: or others, they will reveal just the amount of money,
• 7:21
  Speaker 3: the amount of individuals they have focused only on deterring
• 7:26
  Speaker 3: cybersecurity risks. On a given day, there's just they paint
• 7:29
  Speaker 3: a picture of an onslaught on the bank's security systems
• 7:33
  Speaker 3: on their websites daily. That's just been a fact of
• 7:36
  Speaker 3: life for years. Shouldn't they be on top of it
• 7:38
  Speaker 3: by now?
• 7:39
  Speaker 2: I think part of the thing is they feel like
• 7:40
  Speaker 2: they have been spending a lot of time on this
• 7:42
  Speaker 2: and this has still caught them somewhat unaware, even though
• 7:45
  Speaker 2: they knew there were cyber risks. I think it's the escalation,
• 7:48
  Speaker 2: the fact that this could all happen so quickly, and
• 7:50
  Speaker 2: I think some of them have been somewhat blindsided by
• 7:53
  Speaker 2: this because certainly there is a lot of focus and
• 7:56
  Speaker 2: has been for a long time, and people have been
• 7:58
  Speaker 2: warning about cyber risk, but it's hard to get people
• 8:01
  Speaker 2: to really understand it until it becomes a real and
• 8:05
  Speaker 2: a present danger. So there were some similar conversations going
• 8:07
  Speaker 2: on around the quantum computing capacity, and people had been
• 8:10
  Speaker 2: concerned that when we have holescale quantum, that's going to
• 8:13
  Speaker 2: lead to a great escalation around cyber attacks, and people
• 8:16
  Speaker 2: were thinking about that talking about that, but it was
• 8:18
  Speaker 2: seen as something which is going to become a reality
• 8:20
  Speaker 2: in five ten years. So I think the idea that
• 8:23
  Speaker 2: there would be this step change in cybersecurity demands so
• 8:26
  Speaker 2: quickly is what's called people unawarees. And also the financial
• 8:30
  Speaker 2: system hasn't got very good tools to deal with this
• 8:33
  Speaker 2: in a harmonized, unified way, and that's also been a
• 8:37
  Speaker 2: feature of some of the conversations this week.
• 8:39
  Speaker 3: And Michael, even as we've had this conversation, and I
• 8:42
  Speaker 3: noticed that even in the coverage over these last few days,
• 8:45
  Speaker 3: we had the story that I mentioned at the top
• 8:47
  Speaker 3: about Treasury Secretary bushering in banks, and it was understood
• 8:52
  Speaker 3: to be him warning about this model, but actually it
• 8:55
  Speaker 3: turns out it was also about encouraging banks to use
• 8:59
  Speaker 3: this model to identify the risks and the holes in
• 9:04
  Speaker 3: their system ahead of others, because it's so good at
• 9:08
  Speaker 3: identifying weaknesses. So I guess it does raise this question
• 9:11
  Speaker 3: is it a threat or is it actually something that's
• 9:13
  Speaker 3: going to make us safer this kind of model.
• 9:16
  Speaker 1: So absolutely both, and I think a lot of it
• 9:18
  Speaker 1: depends on the timing itself. So in the immediate short term,
• 9:21
  Speaker 1: it's much more of a threat, just because things are
• 9:23
  Speaker 1: changing so quickly, and the velocity, the speed of the
• 9:26
  Speaker 1: threat environment has accelerated so dramatically, and so one of
• 9:29
  Speaker 1: the things is for the attackers have always had a
• 9:32
  Speaker 1: structural advantage, let's say pre AI, in the sense that
• 9:34
  Speaker 1: they could choose a time and place where they attack
• 9:37
  Speaker 1: and they only need to succeed once. And defenders, as
• 9:39
  Speaker 1: you already alluded to banks for example, face huge campaigns
• 9:42
  Speaker 1: of attacks all over the place and they need to
• 9:43
  Speaker 1: succeed one hundred percent of the time, and if they
• 9:45
  Speaker 1: fail even once, then consequences could be catastrophic. What's changed
• 9:49
  Speaker 1: with anthropic with MYTHOS now is that you see this
• 9:52
  Speaker 1: general upskilling of attacking capabilities across the board.
• 9:56
  Speaker 4: Right.
• 9:56
  Speaker 1: That's the initial concern, which is that many people who
• 9:58
  Speaker 1: didn't have the capabilities to attack top institution cyber defense
• 10:01
  Speaker 1: teams could now use Mythos, assuming they get access to
• 10:04
  Speaker 1: it in a year or so when this becomes public,
• 10:06
  Speaker 1: to then attack institution that a scale we've never known before.
• 10:09
  Speaker 1: But in the long term, the reason this potentially could
• 10:11
  Speaker 1: actually be a benefit is again as you discussed, there's
• 10:14
  Speaker 1: huge potential for this to help defenders scan for vulnerabilities
• 10:18
  Speaker 1: and close the sort of resource gap that they had
• 10:21
  Speaker 1: with the attacking side. There was always a scale mismatch
• 10:24
  Speaker 1: between attackers and defenders, but this helps close that scale
• 10:27
  Speaker 1: for the defender side, who also have an internal information advantage.
• 10:30
  Speaker 1: They've always had this advantage as well, but were just
• 10:32
  Speaker 1: never able to fully exploit it in the sense that
• 10:34
  Speaker 1: they know their own systems the best, and once they
• 10:37
  Speaker 1: have the scale and the resources that Mythos provides them
• 10:40
  Speaker 1: to act on that information advantage, it actually means we
• 10:42
  Speaker 1: could get mount for secure systems. They could find all
• 10:44
  Speaker 1: the vulnerabilities, gradually patched them out over time, and once
• 10:48
  Speaker 1: you patch a vulnerability, it's gone, so an attacker can
• 10:51
  Speaker 1: no longer use that. So in the long run you
• 10:52
  Speaker 1: could see this certainly assuming banks and top institutions act
• 10:56
  Speaker 1: quickly enough to look at their own vulnerabilities, which is
• 10:59
  Speaker 1: what this anthropics headstart is giving them right now, then
• 11:02
  Speaker 1: you could see this really benefiting the defense.
• 11:04
  Speaker 3: There's often a discussion around economists saying that things will
• 11:07
  Speaker 3: be fine in the long run, and then you know,
• 11:10
  Speaker 3: on the famous Canes quote about when the.
• 11:12
  Speaker 4: Long run, we're all dead.
• 11:13
  Speaker 3: In the short term, it sounds like if you think
• 11:16
  Speaker 3: the period of quite a heightened risk is going to
• 11:19
  Speaker 3: be when these models are out there, but it's not
• 11:21
  Speaker 3: clear in whose hands they're in, and the state of
• 11:25
  Speaker 3: knowledge on the side of the banks is perhaps mixed.
• 11:28
  Speaker 3: It seems like we're going to have quite a few
• 11:30
  Speaker 3: bumps on the road to that more secure future that
• 11:33
  Speaker 3: you just painted.
• 11:34
  Speaker 1: Yes, absolutely, and even that more secure future is conditional
• 11:37
  Speaker 1: on a lot of this AI being properly used and
• 11:40
  Speaker 1: properly implemented and deployed across most institutions. One of the
• 11:44
  Speaker 1: things I just discussed about defenders having an advantage is only
• 11:47
  Speaker 1: true in the immediate short term for top institutions with
• 11:50
  Speaker 1: access to these models, with the budgets and the cybersecurity
• 11:53
  Speaker 1: teams to properly defend their systems and networks. A lot
• 11:57
  Speaker 1: of these smaller bank, smaller institutions, smaller firms across the
• 12:00
  Speaker 1: boardroom in many cases even now, don't have proper cybersecurity
• 12:03
  Speaker 1: teams are going to be even more exposed and vulnerable
• 12:05
  Speaker 1: to heightened campaigns from attackers with much more skill and resources,
• 12:10
  Speaker 1: assuming that they do get access to mythos in time.
• 12:13
  Speaker 1: And in part what needs to happen is I think
• 12:16
  Speaker 1: the infrastructure providers for the Internet for the financial system.
• 12:20
  Speaker 1: The top institutions themselves need to be much more proactive
• 12:22
  Speaker 1: about pushing out updates and patches to make sure that
• 12:25
  Speaker 1: most of these smaller institutions aren't running their own infrastructure right,
• 12:28
  Speaker 1: and so the infrastructure behind them needs to be more
• 12:30
  Speaker 1: proactive and making sure vulnerabilities are fully addressed so that
• 12:33
  Speaker 1: as a network effect, the totality of the system and
• 12:36
  Speaker 1: the economy is protected. And that's what Anthropic is doing
• 12:38
  Speaker 1: with Project glass Wing. In addition to their twelve launch partners,
• 12:41
  Speaker 1: these top banks and institutions, they've added in forty plus
• 12:44
  Speaker 1: infrastructure providers because they know this is the quickest and
• 12:47
  Speaker 1: fastest way to reach the most people.
• 12:49
  Speaker 4: Laura, like the regulators, you're learning on the fly.
• 12:52
  Speaker 3: You're you're not from our tech team, but I suspect
• 12:55
  Speaker 3: you're going to get more and more by the hour.
• 12:57
  Speaker 4: You're getting better informed on this.
• 12:59
  Speaker 3: But just on that point of the sort of this
• 13:02
  Speaker 3: double edged aspect where you've got to be very frightened
• 13:06
  Speaker 3: of this model, but you actually want to make sure
• 13:08
  Speaker 3: that your own institutions get it first so that they
• 13:11
  Speaker 3: can actually use it to spot vulnerabilities. The regulators that
• 13:17
  Speaker 3: you listen to are they getting their heads around that.
• 13:20
  Speaker 4: I noticed that we had a store.
• 13:21
  Speaker 3: You know, it does seem like some UK financial firms
• 13:23
  Speaker 3: are going to get hold of this model.
• 13:26
  Speaker 4: How is that playing?
• 13:27
  Speaker 2: So it really has opened up an interesting debate about
• 13:30
  Speaker 2: who gets access and who doesn't and also who decides,
• 13:33
  Speaker 2: because it's not actually for the FED, for the Battle
• 13:35
  Speaker 2: Committee for the Bank of England to decide. This model
• 13:38
  Speaker 2: is owned by a non state actor, so they can encourage,
• 13:43
  Speaker 2: but they don't really have a role there. And the
• 13:45
  Speaker 2: banks who are not currently able to access it, which
• 13:49
  Speaker 2: is every win aside from the biggest US banks, they
• 13:51
  Speaker 2: do feel like they are at a disadvantage because they
• 13:54
  Speaker 2: can't plan and they can't prepare, and they're very keen
• 13:57
  Speaker 2: to get access to it. There are some of the
• 13:59
  Speaker 2: European officials who were here this week who are trying
• 14:02
  Speaker 2: to softly encourage their peers in the US to encourage
• 14:06
  Speaker 2: the US technology company which owns the model to give
• 14:08
  Speaker 2: the European banks access, but there's no obvious process for that.
• 14:12
  Speaker 2: So it's one of those things whereas this tech evolves,
• 14:16
  Speaker 2: we may need to have a more standardized process for
• 14:18
  Speaker 2: if there is a systemically important development in the tech,
• 14:22
  Speaker 2: there's an actual process for deciding who gets access. It's
• 14:25
  Speaker 2: done in a more democratic way, in a more equal way,
• 14:28
  Speaker 2: and that's a hard thing to happen. And one of
• 14:29
  Speaker 2: the challenges for Europe is Europe hasn't exactly got a
• 14:33
  Speaker 2: version in tech industry itself so exta that there are
• 14:36
  Speaker 2: AI issues, they are likely to involve non European companies,
• 14:41
  Speaker 2: and that means that the European large companies who are
• 14:43
  Speaker 2: effectually clients of these AI firms are not going to
• 14:47
  Speaker 2: be at the front of the queue when it comes
• 14:48
  Speaker 2: to getting assistance to dealing with these issues, and that
• 14:51
  Speaker 2: is a challenge which Europe is very acutely aware of.
• 15:01
  Speaker 3: Michael, I mean, it's odd because Anthropic, you know, we
• 15:04
  Speaker 3: had this news about Scott Bessen. But of course the
• 15:07
  Speaker 3: main dynamic between Anthropic and the US administration recently has
• 15:11
  Speaker 3: been around the arguments with the Defense Department and being categorized,
• 15:16
  Speaker 3: whether it was legal or not, being categorized as a
• 15:19
  Speaker 3: supply chain threat.
• 15:20
  Speaker 4: How should we read that.
• 15:22
  Speaker 3: That's the same administration that has said it's a supply
• 15:26
  Speaker 3: chain threat has encouraged all of its contractors to stop
• 15:29
  Speaker 3: using Anthropic Tech is also encouraging some financial firms precisely
• 15:35
  Speaker 3: to use this tool. Have they just given up on
• 15:38
  Speaker 3: the supply chain threat aspect? I, as.
• 15:40
  Speaker 1: Well as many others, would find the supply chain is
• 15:42
  Speaker 1: designation questionable for Anthropic and that.
• 15:45
  Speaker 4: The administration hasn't quite fescin.
• 15:47
  Speaker 1: Yes, the specific use of Nthropic and the argument over
• 15:50
  Speaker 1: that in that context is very much about the Department
• 15:53
  Speaker 1: of Defense not wanting to be constrained in any way
• 15:55
  Speaker 1: in the use of entropics model. The context for its
• 15:58
  Speaker 1: wider use in the financial system and protecting many of
• 16:02
  Speaker 1: our top institutions is much more about the capability of
• 16:04
  Speaker 1: the model itself. It's a difference between scope of responsibility
• 16:08
  Speaker 1: versus raw capability, and so in this case, I think
• 16:11
  Speaker 1: even even if you assume the supply chain risk designation
• 16:14
  Speaker 1: was somehow entirely legitimate, you would still have to be
• 16:17
  Speaker 1: careful about let's say a Chinese open weight model that's
• 16:20
  Speaker 1: suddenly had this capability. You would still see Treasury briefing
• 16:23
  Speaker 1: all the top banks and institutions to make sure are
• 16:25
  Speaker 1: you aware of this threat? Are you preparing it with
• 16:27
  Speaker 1: that this threat? Are you actually using this to defend
• 16:29
  Speaker 1: your own system?
• 16:29
  Speaker 4: It's not just the warning about it, is encouraging them
• 16:32
  Speaker 4: to use it.
• 16:34
  Speaker 1: Yes, Yeah, I wouldn't be the first to point out
• 16:36
  Speaker 1: that there's a lot of contradictions with this administration.
• 16:38
  Speaker 3: Well, although going back to what Laura said, we may
• 16:41
  Speaker 3: or may not the way that Pete Hexa sounds or
• 16:43
  Speaker 3: some of the arguments that were used by the Pentagon,
• 16:45
  Speaker 3: but the point of principle that a private sector firms
• 16:49
  Speaker 3: shouldn't be the one dictating where these powers sit and
• 16:53
  Speaker 3: how they're used. That is something that I think people
• 16:57
  Speaker 3: across the spectrum would have some sympathy with. If this
• 17:01
  Speaker 3: is something that is the difference between having an enormous
• 17:06
  Speaker 3: cyber risk to household deposits important systemic institutions in the
• 17:13
  Speaker 3: financial system, if it's a private sector company that's determining
• 17:17
  Speaker 3: whether you have protection against that or not. There it
• 17:20
  Speaker 3: does seem something undemocratic and wrong about that.
• 17:23
  Speaker 1: Yeah, absolutely, And I think this is a conversation that's
• 17:25
  Speaker 1: going to intensify as other firms catch up to war
• 17:28
  Speaker 1: Anthropic is OpenAI a week later released GPT five. Four
• 17:32
  Speaker 1: cybursts insert themselves in the conversation to and prove that
• 17:35
  Speaker 1: they have an equally capable model equally scary model. I
• 17:39
  Speaker 1: think as these capabilities progress, you're going to see more
• 17:41
  Speaker 1: and more discussion about potential government intervention, even talk about
• 17:46
  Speaker 1: government stakes or nationalization to some degree, because these capabilities
• 17:50
  Speaker 1: are game changing that there's no way that a government
• 17:52
  Speaker 1: would tolerate all this power to be held entirely in
• 17:54
  Speaker 1: the hands of private actors. The US right now is
• 17:57
  Speaker 1: by far the most lenient on AI regular and accommodation
• 18:01
  Speaker 1: of private AI interests. But even that I don't think
• 18:04
  Speaker 1: can hold over time because we're not seeing a plateau
• 18:06
  Speaker 1: on I capabilities anytime soon. And consider the difference between
• 18:09
  Speaker 1: mythos now and where GPT was a year ago, and
• 18:12
  Speaker 1: then build that out two or three more years where
• 18:14
  Speaker 1: we might be with AI and with there being no
• 18:17
  Speaker 1: seeming plateau like technical plateau on the horizon. I think
• 18:20
  Speaker 1: that's really a conversation that needs to be had.
• 18:22
  Speaker 3: And Laura, you've raised the point about Europe. You listen
• 18:25
  Speaker 3: to what Michael just said, and then you listen to
• 18:29
  Speaker 3: conversations with European policy makers saying we can't afford to
• 18:32
  Speaker 3: have all of our technology stack everything be dependent on
• 18:36
  Speaker 3: US firms. In this case, that's a lost cause. Presumably,
• 18:40
  Speaker 3: if you're talking about the security of the financial system,
• 18:43
  Speaker 3: they're absolutely going to have to be dependent on the
• 18:46
  Speaker 3: US forever. More.
• 18:47
  Speaker 2: It seems like not just on the US, so I
• 18:49
  Speaker 2: think in this contact you're the US, we also have Asia,
• 18:52
  Speaker 2: so we also have the Chinese AI companies, but it's
• 18:55
  Speaker 2: all foreign And to be honest, the Asia stuff is
• 18:57
  Speaker 2: actually scarier for the policy makers because while you're banks,
• 19:00
  Speaker 2: they're not at the start at the top of the queue.
• 19:02
  Speaker 2: For the US, there's at least friendly and open relationships.
• 19:05
  Speaker 2: There is more concern that if there were a company
• 19:08
  Speaker 2: in a less collaborative jurisdiction whose which actually developed this
• 19:13
  Speaker 2: kind of technology, they might not give Western banks a
• 19:16
  Speaker 2: heads up to get their defenses out. And I think
• 19:17
  Speaker 2: that's the big concern is what if this develops somewhere
• 19:21
  Speaker 2: that we don't have at least a line of contact
• 19:24
  Speaker 2: into to try and get our banks defended. But in
• 19:27
  Speaker 2: terms of Europe, the ship has sailed. There are aspirational
• 19:30
  Speaker 2: projects around let's try and invest a lot of money
• 19:32
  Speaker 2: and build our own tech giants, but really that is
• 19:35
  Speaker 2: a very long term plan and most people would say
• 19:37
  Speaker 2: that's probably never going to come to fruition so what's
• 19:39
  Speaker 2: about for them is trying to find the best oversight
• 19:43
  Speaker 2: model for the technology that we live in. And that
• 19:46
  Speaker 2: is a challenging conversation because ultimately it does come to
• 19:49
  Speaker 2: like the main interest of the US financial regulators is
• 19:53
  Speaker 2: protecting the US financial system. It's the US system first
• 19:57
  Speaker 2: and then the global system second. That is the way
• 20:00
  Speaker 2: people increasingly think about the financial systems of the world.
• 20:04
  Speaker 2: There has been a real fragmentation in terms of how
• 20:07
  Speaker 2: people see the global financial system, and in these kind
• 20:10
  Speaker 2: of moments, people do think of their own industries first,
• 20:13
  Speaker 2: and I think only once they have secured the biggest,
• 20:16
  Speaker 2: the main US interests, then they will expand the efforts
• 20:20
  Speaker 2: to their colleagues in Europe and to other places which
• 20:23
  Speaker 2: are key markets for them.
• 20:24
  Speaker 3: Yeah, and I was struck by a piece that you've
• 20:26
  Speaker 3: written where you say officials privately lament the every man
• 20:29
  Speaker 3: for himself approach that's increasingly creeping into global conversations on
• 20:33
  Speaker 3: finance and could hinder the effective sharing of information on
• 20:38
  Speaker 3: risks globally. It feels a bit like the issue we
• 20:41
  Speaker 3: had with vaccine development after COVID, that you had individual
• 20:46
  Speaker 3: countries really looking to protect their own citizens before anyone
• 20:50
  Speaker 3: else got protected. But is that plausible in a very
• 20:53
  Speaker 3: globally interconnected financial system, doesn't Don't America have a stake
• 20:57
  Speaker 3: in europe financial system, being say, for given how much
• 21:01
  Speaker 3: the inter relationships there are.
• 21:03
  Speaker 2: It certainly has a stake. So I think it is
• 21:04
  Speaker 2: a different dynamic to vaccines in that sense. But what
• 21:06
  Speaker 2: I would say is it definitely has an interest in
• 21:09
  Speaker 2: global financial stability, but it has an even bigger interest
• 21:13
  Speaker 2: in doing what it can to protect the US financial systems.
• 21:16
  Speaker 2: I think there's certainly a hierarchy there. It's not that
• 21:18
  Speaker 2: the global financial system isn't at all relevant to the US,
• 21:21
  Speaker 2: It's just the main thing, and all of the domestic
• 21:23
  Speaker 2: agenda here is about trying to protect the US. We
• 21:27
  Speaker 2: hear it in terms of the debate around growth, around competitiveness.
• 21:31
  Speaker 2: We hear it in the EU as well. People's primary
• 21:33
  Speaker 2: concern is trying to help their own industries first, and
• 21:36
  Speaker 2: then we'll deal with the global picture.
• 21:38
  Speaker 3: Michael Love their responses that regulators should have that address
• 21:43
  Speaker 3: directly some of the things you've been talking about in
• 21:46
  Speaker 3: terms of the key thing being not just access to
• 21:50
  Speaker 3: these kind of tools, but the incentives that institutions would
• 21:54
  Speaker 3: have to patch the flaws that they identify. At the moments,
• 21:58
  Speaker 3: banks aren't forced to do that, but just trusting that
• 22:01
  Speaker 3: they will put in place things that will protect their
• 22:03
  Speaker 3: interest Do you think that's something that needs to change.
• 22:06
  Speaker 3: Regulators will need to change the way they look at it.
• 22:08
  Speaker 1: I actually do think there's potentially a case to be
• 22:11
  Speaker 1: made for stronger regulation in one consideration, because even when
• 22:14
  Speaker 1: you have a lot of these banks protecting themselves out
• 22:16
  Speaker 1: of their own interests, it's entirely possible for some of
• 22:19
  Speaker 1: them to have a greater risk appetite over a cybersecurity
• 22:21
  Speaker 1: than others. And you can see that in just basic
• 22:23
  Speaker 1: financial transactions all the time with certain banks, and.
• 22:26
  Speaker 3: You can see that with the glibl functually crisis. So
• 22:29
  Speaker 3: a lot of them didn't protect themselves as much as
• 22:31
  Speaker 3: the FED and others anticipated.
• 22:33
  Speaker 1: Yes, exactly, And I do think there needs to be
• 22:35
  Speaker 1: more uniform standard that ensures that banks are properly addressing
• 22:40
  Speaker 1: cybersecurity vulnerabilities and they're up to date and following whatever
• 22:44
  Speaker 1: the latest patching schemes or general paradigms are in the
• 22:48
  Speaker 1: way that a lot of other critical infrastructure industries are
• 22:50
  Speaker 1: as well. Right now, it just seems to be a
• 22:52
  Speaker 1: dialogue and a partnership, but I think the government does
• 22:54
  Speaker 1: probably need to take a more active role as these
• 22:56
  Speaker 1: models get more powerful.
• 22:57
  Speaker 3: I have heard people talk about different ways of thinking
• 23:01
  Speaker 3: about this risk. There's having the absolute best tools and
• 23:04
  Speaker 3: being part of an arms race where you just make
• 23:07
  Speaker 3: sure that you've got a better tool than the hackers have.
• 23:09
  Speaker 3: But there is also the idea that you limit the
• 23:13
  Speaker 3: risks of any given threat blowing up the rest of
• 23:18
  Speaker 3: the system. I think of it in a very kind
• 23:20
  Speaker 3: of analog way, as you unplug the computer, but you
• 23:23
  Speaker 3: know that you isolate. You have systems that can be
• 23:25
  Speaker 3: isolated from global networks, so that when you assume you
• 23:29
  Speaker 3: are going to have attax and you just make sure
• 23:30
  Speaker 3: they can't spread through the system. Is there an aspect
• 23:32
  Speaker 3: of that, because it seems like our whole conversation is
• 23:34
  Speaker 3: more about the arms race. You're just going to have
• 23:36
  Speaker 3: to keep up and make sure that you're getting all
• 23:38
  Speaker 3: of your updates all of the time, which for so
• 23:40
  Speaker 3: many institutions is not going to be practical.
• 23:42
  Speaker 1: I think would be very difficult to do that because
• 23:45
  Speaker 1: in many ways, that's effectively the opposite of how the
• 23:48
  Speaker 1: financial system and the global economy works, and even how
• 23:51
  Speaker 1: Internet infrastructure works. Like I talk about the concentration of
• 23:55
  Speaker 1: core being processing vendors, I also talk about the concentration
• 23:57
  Speaker 1: of cryptographic protocols that protect all the fine networks. There's
• 24:01
  Speaker 1: just not really an effective, scalable way to wall off
• 24:04
  Speaker 1: individual systems. I'm sure you could do that for individual
• 24:07
  Speaker 1: niche systems, but the moment that a system needs to
• 24:10
  Speaker 1: talk with another system, that inevitably exposes it to all
• 24:13
  Speaker 1: the network vulnerabilities that come with it. And that's the
• 24:15
  Speaker 1: tragedy of the Internet in itself, is that this kind
• 24:18
  Speaker 1: of system of connections was never built for security first,
• 24:20
  Speaker 1: and we're essentially patching in security and resilience on top
• 24:23
  Speaker 1: of that in the best way we can.
• 24:24
  Speaker 3: And on that bombshell, Laura, last question to you. We
• 24:26
  Speaker 3: thought we would be probably talking about AI this week
• 24:29
  Speaker 3: at the Well Bank and IMF meetings, but possibly more
• 24:33
  Speaker 3: around risks to jobs and economy's longer term but short
• 24:36
  Speaker 3: term potentially the risk to the economy from an AI
• 24:40
  Speaker 3: bubble bursting. Do you think for some of the officials
• 24:44
  Speaker 3: you're talking to, the AI related risk that actually could
• 24:48
  Speaker 3: be most threatening to the global economy right now, the
• 24:51
  Speaker 3: financial system is the one we've been talking about.
• 24:53
  Speaker 2: It's certainly the one that they're most concerned about this week,
• 24:56
  Speaker 2: but it's also the recency bias. This is something that
• 24:58
  Speaker 2: just emerged a couple of days before the IMF meetings,
• 25:00
  Speaker 2: and everyone's talking about thattand it. Well, it's really been
• 25:05
  Speaker 2: interesting watching people who always understand what they're talking about
• 25:07
  Speaker 2: begin the conversation by asking me what I've heard, because
• 25:10
  Speaker 2: it's all the fact finding mission this week. But I
• 25:12
  Speaker 2: think there have been long running concerns about the credit
• 25:15
  Speaker 2: quality around AI, the fact there's been a lot of
• 25:17
  Speaker 2: money pumped into data centers. How is that going to
• 25:20
  Speaker 2: play out in the longer term. There's also the jobs
• 25:22
  Speaker 2: macroeconomic thing, but those questions they are also somewhat unanswerable.
• 25:26
  Speaker 2: So there is a sense that we're going to get
• 25:27
  Speaker 2: a newer term resolution to how this issue plays out
• 25:31
  Speaker 2: than the long term impact of what AI is going
• 25:33
  Speaker 2: to mean for jobs, economics and for the bigger world.
• 25:37
  Speaker 2: That's going to take longer to play out. And similarly,
• 25:39
  Speaker 2: we could see the AI bubble. The AI bubble has
• 25:42
  Speaker 2: been warned about for quite some time and has proved
• 25:44
  Speaker 2: pretty resilient, so far because there's only so many ways
• 25:48
  Speaker 2: that people can frame the same fears about the AI bubble.
• 25:51
  Speaker 2: So what I think is has given this piece more
• 25:54
  Speaker 2: oxygen is also the fact that people have new things
• 25:57
  Speaker 2: to say on it. Even if the new thing that
• 25:59
  Speaker 2: they're saying is this sounds like it could be very
• 26:01
  Speaker 2: bad and we don't know what's going to happen, but
• 26:03
  Speaker 2: it is at least something new to say. We've heard
• 26:06
  Speaker 2: a lot of warnings about the AI bubble. We've heard
• 26:08
  Speaker 2: a lot of warnings about AI and jobs.
• 26:10
  Speaker 3: Okay, so enormous amounts of money and attention and potentially
• 26:14
  Speaker 3: large chunks of the financial system going into something that
• 26:17
  Speaker 3: we don't fully understand what could possibly go wrong. Thank
• 26:20
  Speaker 3: you very much, Laura and Michael, thank you, Thank you
• 26:30
  Speaker 3: thanks for listening to Trumponomics from Bloomberg. It was hosted
• 26:32
  Speaker 3: by me Stephanie Flanders. I was joined by Laura Noonan
• 26:35
  Speaker 3: of Bloomberg News and Michael Dunn from Bloomberg Economics. Trumpnomics
• 26:39
  Speaker 3: was produced by Summersadi and Moses and with help from
• 26:43
  Speaker 3: Rachel Lewis Christy and Amy Keen. Sound design was by
• 26:47
  Speaker 3: Blake Naples and Kelly Gary And to help others find
• 26:50
  Speaker 3: this and enjoy it, rate.
• 26:53
  Speaker 4: And review it highly. Wherever you listen to podcasts,