1 00:00:01,440 --> 00:00:04,160 Speaker 1: The European Union is on the brink of becoming the 2 00:00:04,320 --> 00:00:08,440 Speaker 1: first major power in the world to regulate artificial intelligence. 3 00:00:08,520 --> 00:00:11,319 Speaker 2: Politicians in Europe will vote on a proposal to bring 4 00:00:11,360 --> 00:00:14,240 Speaker 2: in a law that would govern the use of artificial intelligence. 5 00:00:14,400 --> 00:00:18,760 Speaker 3: I think we have made history today. We have set 6 00:00:18,920 --> 00:00:23,680 Speaker 3: the way for the dialogue that we will need to 7 00:00:23,760 --> 00:00:26,040 Speaker 3: have we started having with the rest of the world. 8 00:00:26,120 --> 00:00:35,880 Speaker 3: Or now we can build responsible AI for our globe. 9 00:00:37,000 --> 00:00:40,559 Speaker 4: I'm Roslin Mathieson in for Wis Kosovo today on the 10 00:00:40,560 --> 00:00:43,559 Speaker 4: big take. What's the European Union trying to do to 11 00:00:43,680 --> 00:00:45,360 Speaker 4: regulate artificial intelligence? 12 00:00:52,280 --> 00:00:54,880 Speaker 5: I think if this technology goes wrong, it can go 13 00:00:55,000 --> 00:00:58,480 Speaker 5: quite wrong, and we want to be vocal about that. 14 00:00:58,520 --> 00:01:00,920 Speaker 5: We want to work with the government to prevent that 15 00:01:00,960 --> 00:01:04,000 Speaker 5: from happening. It's one of my areas of greatest concern. 16 00:01:04,560 --> 00:01:07,760 Speaker 4: What you just heard was Sam Altman, the CEO of OpenAI, 17 00:01:08,240 --> 00:01:12,360 Speaker 4: the artificial intelligence research lab that created the generative AI 18 00:01:12,400 --> 00:01:16,440 Speaker 4: model known as CHATGBT. Earlier this year, he testified in 19 00:01:16,480 --> 00:01:19,280 Speaker 4: front of the US Senate calling for more regulation of 20 00:01:19,319 --> 00:01:22,280 Speaker 4: AI tech in both his own company and others, and 21 00:01:22,440 --> 00:01:25,440 Speaker 4: just last week, open Ai, along with Google and Microsoft, 22 00:01:25,520 --> 00:01:28,959 Speaker 4: went one step further they're creating an industry watchdog to 23 00:01:29,000 --> 00:01:32,319 Speaker 4: make sure their AI models are safe. While regulation is 24 00:01:32,360 --> 00:01:35,479 Speaker 4: moving at a snail's pace in the US. The European 25 00:01:35,600 --> 00:01:38,240 Speaker 4: Union has set out to be the first Western administration 26 00:01:38,400 --> 00:01:42,200 Speaker 4: to create a comprehensive set of guardrails around AI, known 27 00:01:42,240 --> 00:01:46,479 Speaker 4: as the AI Act. But why are European regulations so important? 28 00:01:47,440 --> 00:01:52,080 Speaker 2: If you wanted to evade the European regulations and, for instance, 29 00:01:52,200 --> 00:01:55,960 Speaker 2: offer them this same AI product somewhere in the United States, 30 00:01:56,000 --> 00:01:59,120 Speaker 2: you would need to retrain the model and carve out 31 00:01:59,320 --> 00:02:02,680 Speaker 2: all the European data from that model, so that is 32 00:02:02,800 --> 00:02:04,960 Speaker 2: often unappealing for companies. 33 00:02:05,400 --> 00:02:08,359 Speaker 4: That was our new Bradford Law, professor at Columbia Law 34 00:02:08,400 --> 00:02:11,240 Speaker 4: School and author of the term the Brussels effect, the 35 00:02:11,320 --> 00:02:13,840 Speaker 4: idea that what is agreed in the EU becomes the 36 00:02:13,880 --> 00:02:17,600 Speaker 4: global standard for regulation. We'll hear more from Arnu later 37 00:02:17,680 --> 00:02:21,480 Speaker 4: in the episode, but first, Bloomberg reported, Gillian Deutsch joins 38 00:02:21,520 --> 00:02:23,600 Speaker 4: us from Brussels to tell us a bit more about 39 00:02:23,639 --> 00:02:25,400 Speaker 4: what exactly the AI Act is. 40 00:02:27,040 --> 00:02:30,160 Speaker 6: The EU'SAI Act is really the first attempt by a 41 00:02:30,200 --> 00:02:35,320 Speaker 6: Western government to comprehensively regulate artificial intelligence, and the European 42 00:02:35,320 --> 00:02:38,079 Speaker 6: Commission first proposers back in twentyenty one, so this is 43 00:02:38,240 --> 00:02:42,160 Speaker 6: pre chagipt hype, and their approach was really to regulate 44 00:02:42,200 --> 00:02:45,640 Speaker 6: the use of the technology, not the technology itself. So 45 00:02:45,680 --> 00:02:48,720 Speaker 6: they called it a risk based approach, and they prohibited 46 00:02:48,760 --> 00:02:51,800 Speaker 6: certain kinds of technology or a certain kinds of uses 47 00:02:51,840 --> 00:02:55,320 Speaker 6: of AI. For example, they've banned social scoring, and they 48 00:02:55,360 --> 00:02:58,160 Speaker 6: often cite China and this kind of idea that you'd 49 00:02:58,200 --> 00:03:01,040 Speaker 6: have a social score based on your behavior as a citizen. 50 00:03:01,360 --> 00:03:03,480 Speaker 6: That's one of the things a're trying to avoid. And 51 00:03:03,560 --> 00:03:06,200 Speaker 6: most of the puzzle focused on so called high risk 52 00:03:06,320 --> 00:03:06,960 Speaker 6: use cases. 53 00:03:07,400 --> 00:03:09,400 Speaker 7: So this is if law enforcement. 54 00:03:08,960 --> 00:03:11,959 Speaker 6: Wants to use AI, or a company wants to use 55 00:03:12,160 --> 00:03:16,080 Speaker 6: artificial intelligence to scan employment applications, for example, they want 56 00:03:16,120 --> 00:03:19,000 Speaker 6: to make sure that is overseen by human There are 57 00:03:19,000 --> 00:03:21,640 Speaker 6: certain assessments to make sure that their technology is not 58 00:03:21,680 --> 00:03:25,480 Speaker 6: misused or enforces bias. But the vast majority of AI 59 00:03:25,680 --> 00:03:28,840 Speaker 6: at that point would be unregulated if it's not using 60 00:03:28,880 --> 00:03:32,200 Speaker 6: those high risk circumstances. So we didn't really see actually 61 00:03:32,200 --> 00:03:34,120 Speaker 6: anything explicitly on gurnet of AI. 62 00:03:34,160 --> 00:03:36,440 Speaker 7: If you look for the terms like large. 63 00:03:36,240 --> 00:03:40,120 Speaker 6: Language models or the fundation models, these kinds of things 64 00:03:40,120 --> 00:03:42,280 Speaker 6: that we're talking about so much now but not really 65 00:03:42,320 --> 00:03:45,640 Speaker 6: found in that first draft. It's actually all about this 66 00:03:45,800 --> 00:03:49,600 Speaker 6: high risk use cases. There are some basic transparency requirements 67 00:03:49,640 --> 00:03:52,960 Speaker 6: for deep fakes and chatbots, but that's about it. 68 00:03:53,960 --> 00:03:56,120 Speaker 4: This talk in a bit more detail about what this 69 00:03:56,240 --> 00:03:59,680 Speaker 4: act actually does. Who is it trying to regulate, how 70 00:03:59,720 --> 00:04:01,240 Speaker 4: is it trying to regulate. 71 00:04:02,160 --> 00:04:04,880 Speaker 6: So the Commission's first proposal for the a Act is 72 00:04:04,880 --> 00:04:08,320 Speaker 6: really to focus on the high risk use cases of AI. 73 00:04:08,680 --> 00:04:11,840 Speaker 6: So if a government wants to use, for example, you know, 74 00:04:12,000 --> 00:04:14,640 Speaker 6: artificial intelligence to scan a crowd to look for a 75 00:04:14,720 --> 00:04:17,400 Speaker 6: terrorist or to look for a missing child, the Parliament, 76 00:04:17,440 --> 00:04:20,400 Speaker 6: for example, wants to completely ban that. EU countries, though, 77 00:04:20,440 --> 00:04:23,120 Speaker 6: are saying we need to have some kinds of uses 78 00:04:23,160 --> 00:04:27,040 Speaker 6: for this kind of live facial scanning recognition technology. 79 00:04:27,480 --> 00:04:30,400 Speaker 8: We also are seeing certain companies would be limited in. 80 00:04:30,360 --> 00:04:34,480 Speaker 6: How they use artificial intelligence for sorting through job applications. 81 00:04:34,480 --> 00:04:36,320 Speaker 7: For example, they couldn't really use. 82 00:04:36,160 --> 00:04:38,640 Speaker 6: AI to choose the best candidates, for example, because a 83 00:04:38,640 --> 00:04:41,240 Speaker 6: lot of technology has a long history of choosing white 84 00:04:41,440 --> 00:04:44,800 Speaker 6: candidates over people of color. So those kinds of things 85 00:04:44,800 --> 00:04:47,920 Speaker 6: would be moderated, But the vast majority of AI under 86 00:04:47,960 --> 00:04:50,960 Speaker 6: the current proposal would be allowed or just allowed under 87 00:04:51,200 --> 00:04:54,760 Speaker 6: very loose kinds of controls, like transparency requirements making sure 88 00:04:54,760 --> 00:04:57,080 Speaker 6: people know they're talking to a chatbot, for example, or 89 00:04:57,120 --> 00:04:59,280 Speaker 6: making sure that deep fakes are labeled as such. 90 00:05:00,560 --> 00:05:04,040 Speaker 4: Well, you mentioned that this began really before chat GBT, 91 00:05:04,720 --> 00:05:06,800 Speaker 4: but is that sort of driven some of the more 92 00:05:06,839 --> 00:05:10,080 Speaker 4: recent urgency around it. You said, this is a process 93 00:05:10,120 --> 00:05:12,640 Speaker 4: that began some time ago, but feel like it really 94 00:05:12,640 --> 00:05:13,599 Speaker 4: needs to speed up. 95 00:05:14,200 --> 00:05:17,040 Speaker 6: Yeah, I mean chat chept changed everything here in Brussels. 96 00:05:17,080 --> 00:05:19,640 Speaker 6: So this has been debated for over two years, and 97 00:05:19,720 --> 00:05:23,440 Speaker 6: about a year ago there U countries and lawmakers were 98 00:05:23,440 --> 00:05:26,279 Speaker 6: starting to say there are foundation models that in the 99 00:05:26,279 --> 00:05:29,000 Speaker 6: future could be really powerful, and maybe we should add 100 00:05:29,040 --> 00:05:32,560 Speaker 6: some explicit guardrails around that upcoming technology. But they didn't 101 00:05:32,560 --> 00:05:34,520 Speaker 6: really know what this would look like and they didn't 102 00:05:34,520 --> 00:05:37,880 Speaker 6: know how to regulate it either. So U countries decided, okay, 103 00:05:37,960 --> 00:05:41,080 Speaker 6: let's add some controls over general purpose AI, but they 104 00:05:41,080 --> 00:05:43,320 Speaker 6: didn't really have a great idea for how to do this. 105 00:05:43,720 --> 00:05:46,760 Speaker 6: It was actually lawmakers who were still debating the their 106 00:05:46,880 --> 00:05:50,400 Speaker 6: version of the AI Act who really started adding more 107 00:05:50,400 --> 00:05:54,200 Speaker 6: controls to this because they saw chatjeept explode and become 108 00:05:54,560 --> 00:05:58,040 Speaker 6: from the fastest adopted technologies it'd ever seen, and it 109 00:05:58,080 --> 00:06:00,279 Speaker 6: created this political impetus where they had to add actually 110 00:06:00,279 --> 00:06:03,479 Speaker 6: explicitly regulate CHGBT, Internet of AI for the first time, 111 00:06:03,920 --> 00:06:06,479 Speaker 6: and so they've really come about and added a lot 112 00:06:06,520 --> 00:06:10,480 Speaker 6: more explicit controls. Lawmakers wanted to make sure that companies 113 00:06:10,520 --> 00:06:13,880 Speaker 6: would summarize a copyreaded material that they used train these 114 00:06:13,960 --> 00:06:16,560 Speaker 6: large language models. They also wanted to add some very 115 00:06:16,560 --> 00:06:19,960 Speaker 6: interesting controls companies are not very happy with on making 116 00:06:20,000 --> 00:06:24,080 Speaker 6: sure that they would explain the impact their technologies would have. 117 00:06:24,040 --> 00:06:25,839 Speaker 7: On the environment, unroll of law. 118 00:06:26,120 --> 00:06:28,359 Speaker 6: These are some things that companies argue they don't know 119 00:06:28,400 --> 00:06:30,760 Speaker 6: how to comply with. But it's really the Parliament actually 120 00:06:30,880 --> 00:06:33,640 Speaker 6: going full speed ahead and trying to regulate generative AI 121 00:06:33,680 --> 00:06:34,800 Speaker 6: explicitly for the first time. 122 00:06:35,600 --> 00:06:38,120 Speaker 4: You talk about generative AI, which is really sort of 123 00:06:38,160 --> 00:06:43,960 Speaker 4: AI that basically generates other things audio content, stories at calls. 124 00:06:43,960 --> 00:06:47,760 Speaker 4: Perhaps one day even a podcast could be generated from AI. 125 00:06:48,120 --> 00:06:51,200 Speaker 4: But is that where the concern starts to lie. You 126 00:06:51,240 --> 00:06:53,839 Speaker 4: talk about the focus on high risk in all of this, 127 00:06:54,240 --> 00:06:57,200 Speaker 4: but it just sounds like it's an incredibly complex task 128 00:06:57,360 --> 00:07:00,960 Speaker 4: to try and regulate such as sprawling and fairly new 129 00:07:01,040 --> 00:07:03,800 Speaker 4: and nebulous sector. Is there worry and all of that 130 00:07:03,800 --> 00:07:06,000 Speaker 4: that they step into over regulation? 131 00:07:06,640 --> 00:07:10,040 Speaker 6: Overregulation is a massive concern for everyone. And it's actually 132 00:07:10,080 --> 00:07:12,560 Speaker 6: interesting if you look at France. So France at Yugo 133 00:07:12,640 --> 00:07:15,280 Speaker 6: is actually the country with the spearheading this effort to 134 00:07:15,360 --> 00:07:17,880 Speaker 6: include what they were calling at the time general purpose AI, 135 00:07:18,440 --> 00:07:21,360 Speaker 6: but now a year leader, they're actually they're the ones 136 00:07:21,440 --> 00:07:24,440 Speaker 6: ring alarm bells that we if you do too much regulation, 137 00:07:24,960 --> 00:07:27,360 Speaker 6: we might actually is that you miss out on this 138 00:07:27,480 --> 00:07:30,040 Speaker 6: next wave of technology. And they want to make sure 139 00:07:30,120 --> 00:07:32,760 Speaker 6: that their startups in both in France and both around 140 00:07:32,800 --> 00:07:36,360 Speaker 6: the European Union can actually be leader in this technology. 141 00:07:37,120 --> 00:07:39,640 Speaker 4: And here's what French President and Maliu Macron had to 142 00:07:39,640 --> 00:07:42,160 Speaker 4: say about investment in AI earlier this year. 143 00:07:43,120 --> 00:07:45,000 Speaker 1: I think we're number one in that's not that you are, 144 00:07:45,920 --> 00:07:49,200 Speaker 1: and we have to accelerate. So this is why we 145 00:07:49,240 --> 00:07:51,120 Speaker 1: want to invest more. It's a good thing is that 146 00:07:51,160 --> 00:07:52,800 Speaker 1: we have a lot of good, very good dance. We 147 00:07:52,840 --> 00:07:59,480 Speaker 1: have good matheticians, good that science is a lot of 148 00:07:59,560 --> 00:08:05,000 Speaker 1: dance adapted to this AI environments. We would invest like 149 00:08:05,040 --> 00:08:07,200 Speaker 1: crazy on training and research. 150 00:08:08,720 --> 00:08:10,400 Speaker 4: Is there the sense that the EU is sort of 151 00:08:10,440 --> 00:08:12,520 Speaker 4: the later the front runner on all of this. 152 00:08:13,320 --> 00:08:16,520 Speaker 6: Yeah, it's interesting because you know, we obviously saw Samultman 153 00:08:16,600 --> 00:08:19,360 Speaker 6: from Open AI come out and tell the US Congress 154 00:08:19,400 --> 00:08:23,400 Speaker 6: he wants regulation, and now there's really this impetus in Washington, 155 00:08:23,480 --> 00:08:23,720 Speaker 6: d C. 156 00:08:24,280 --> 00:08:27,400 Speaker 8: To actually regulate AI. But while the US. 157 00:08:27,280 --> 00:08:29,160 Speaker 6: Government is actually looking at this for the first time, 158 00:08:29,440 --> 00:08:32,400 Speaker 6: the is years ahead. Actually they might be ahead though, 159 00:08:32,440 --> 00:08:34,800 Speaker 6: but they still actually haven't figured out just how far 160 00:08:34,880 --> 00:08:37,280 Speaker 6: to go. The EU and US are talking about how 161 00:08:37,320 --> 00:08:39,880 Speaker 6: to regulate AI. The US government is looking at the 162 00:08:39,920 --> 00:08:42,920 Speaker 6: EU government both for as an example of what they 163 00:08:43,000 --> 00:08:45,640 Speaker 6: could do, but also as a cautionary tale of how far, 164 00:08:46,040 --> 00:08:48,079 Speaker 6: I mean too far you could go as a government. 165 00:08:48,520 --> 00:08:50,679 Speaker 6: The US government is even actually lobbied against some of 166 00:08:50,720 --> 00:08:53,040 Speaker 6: these controls as well and said that these are going 167 00:08:53,080 --> 00:08:56,079 Speaker 6: to negatively impact their companies. It's interesting, you do see 168 00:08:56,120 --> 00:08:58,960 Speaker 6: the US government they have some rules in certain states. 169 00:08:59,200 --> 00:09:01,640 Speaker 6: In New York, for example, has laws about how to 170 00:09:01,720 --> 00:09:05,280 Speaker 6: use it in employment situations. But really the EU is 171 00:09:05,400 --> 00:09:09,040 Speaker 6: once again operating as the world's big tech regulator, and 172 00:09:09,080 --> 00:09:11,719 Speaker 6: they have a long reputation of being that world's big 173 00:09:11,760 --> 00:09:12,440 Speaker 6: tech regulator. 174 00:09:12,760 --> 00:09:15,040 Speaker 8: But they also could be a model for going too far. 175 00:09:15,880 --> 00:09:18,840 Speaker 4: Let's talk a little bit about within the EU also 176 00:09:19,000 --> 00:09:23,160 Speaker 4: because there are very big differences between member states. Obviously 177 00:09:23,200 --> 00:09:26,079 Speaker 4: in Brussels itself. You touched on some of these things 178 00:09:26,080 --> 00:09:29,240 Speaker 4: because I imagine for countries who belong to the European Union, 179 00:09:29,679 --> 00:09:31,920 Speaker 4: there's that thing where they want to protect from the 180 00:09:32,000 --> 00:09:35,200 Speaker 4: high risk perspective, but also they want to capture some 181 00:09:35,240 --> 00:09:36,199 Speaker 4: of this business. 182 00:09:36,520 --> 00:09:39,480 Speaker 6: It's interesting because the problem has gone full speed ahead 183 00:09:39,840 --> 00:09:42,040 Speaker 6: and they're adding all these controls like I mentioned, or 184 00:09:42,080 --> 00:09:44,440 Speaker 6: want to add all these controls. The countries are really 185 00:09:44,480 --> 00:09:46,880 Speaker 6: they're aware of all the possible risks. They see that 186 00:09:46,920 --> 00:09:49,520 Speaker 6: this stuff could go too far, but also they want 187 00:09:49,559 --> 00:09:52,920 Speaker 6: to be the hub where startups come and actually, you know, 188 00:09:53,000 --> 00:09:55,800 Speaker 6: build the kind of company that could rival at Google 189 00:09:56,160 --> 00:09:58,160 Speaker 6: or an open AI, and so they do not want 190 00:09:58,200 --> 00:10:01,439 Speaker 6: to overregulate. Even Spain is lead the EU's presidency, they 191 00:10:01,440 --> 00:10:03,760 Speaker 6: have a long reputation of trying to build up their 192 00:10:03,800 --> 00:10:06,760 Speaker 6: AI market and their government has been very much. 193 00:10:06,600 --> 00:10:10,120 Speaker 8: In trying to encourage AI startups to start there. 194 00:10:10,360 --> 00:10:12,520 Speaker 6: So it's really going to be interesting to see how 195 00:10:12,559 --> 00:10:15,880 Speaker 6: far they actually want to pursue rules on CHAGPT. 196 00:10:17,360 --> 00:10:20,920 Speaker 4: My conversation with Jillian continues later in the episode, but 197 00:10:21,160 --> 00:10:24,920 Speaker 4: first will the EU set a precedent for the US. 198 00:10:31,760 --> 00:10:35,080 Speaker 4: I'm joined now by Columbia law professor Arnold Bradford, the 199 00:10:35,120 --> 00:10:38,600 Speaker 4: author of the upcoming book Digital Empires, The Global Battle 200 00:10:38,640 --> 00:10:41,600 Speaker 4: to Regulate Technology. Just to talk a bit more about 201 00:10:41,600 --> 00:10:45,560 Speaker 4: the specifics of this Act in the European context, Arnu, 202 00:10:45,679 --> 00:10:48,120 Speaker 4: I want to start off by asking you how is 203 00:10:48,160 --> 00:10:51,679 Speaker 4: the EU's rights based approach being applied specifically to the 204 00:10:51,720 --> 00:10:52,839 Speaker 4: AI Act. 205 00:10:53,160 --> 00:10:57,480 Speaker 2: So the EU's main concern when it comes to development 206 00:10:57,520 --> 00:11:01,840 Speaker 2: and deployment of AI is how it implicates fundamental rights 207 00:11:01,840 --> 00:11:05,480 Speaker 2: of individuals, So that has been one of the guiding principles. 208 00:11:05,640 --> 00:11:10,120 Speaker 2: So the EU is particularly worried about the individual's fundamental 209 00:11:10,200 --> 00:11:13,680 Speaker 2: right to privacy and the AI can be a very 210 00:11:13,720 --> 00:11:19,320 Speaker 2: powerful tool for surveillance. So the AI Act is looking 211 00:11:19,400 --> 00:11:23,400 Speaker 2: to limit the ways AI systems can be used for 212 00:11:23,600 --> 00:11:28,080 Speaker 2: surveillance purposes, for instance, whether it's predicted policing, but also 213 00:11:28,240 --> 00:11:32,520 Speaker 2: thinking about using facial recognition in public places that can 214 00:11:32,600 --> 00:11:36,800 Speaker 2: potentially then put large segments of the population under surveillance. 215 00:11:37,480 --> 00:11:41,720 Speaker 2: So privacy is one thing, and another particular aspect of 216 00:11:41,760 --> 00:11:45,200 Speaker 2: fundamental rights that is relevant here is discrimination. 217 00:11:45,600 --> 00:11:47,559 Speaker 9: So many companies are using. 218 00:11:47,320 --> 00:11:51,240 Speaker 2: For instance, AI as a tool for recruitment, so the 219 00:11:51,280 --> 00:11:55,600 Speaker 2: AI can determine your access to employment, the AI can 220 00:11:55,640 --> 00:12:00,280 Speaker 2: determine access to education, or banks, financial institutions can use 221 00:12:00,400 --> 00:12:04,520 Speaker 2: AI to screen individuals access to credit, or then the 222 00:12:04,640 --> 00:12:08,199 Speaker 2: states can use it to determine access to public benefits. 223 00:12:08,520 --> 00:12:12,240 Speaker 2: And that is a setting within which some concerns over 224 00:12:12,360 --> 00:12:13,880 Speaker 2: discrimination can arise. 225 00:12:14,640 --> 00:12:18,400 Speaker 4: Just given the complexities of like how you regulate AI 226 00:12:18,600 --> 00:12:21,120 Speaker 4: and the EU trying to hit the right note with 227 00:12:21,280 --> 00:12:24,800 Speaker 4: legislation that's vast and complex, I guess a million dollar 228 00:12:24,920 --> 00:12:27,560 Speaker 4: question is is it even feasible to try and do this? 229 00:12:28,520 --> 00:12:31,720 Speaker 2: So I think that is a really critical question ros 230 00:12:31,800 --> 00:12:35,240 Speaker 2: and I think the EU is conscious of the difficulty 231 00:12:35,240 --> 00:12:37,880 Speaker 2: of the task ahead. But at the same time, I 232 00:12:38,000 --> 00:12:42,480 Speaker 2: don't think it is enough to basically deter the EU 233 00:12:42,600 --> 00:12:46,160 Speaker 2: from going ahead with this legislation. So there's a very 234 00:12:46,559 --> 00:12:49,520 Speaker 2: big temptation, and one often hears that, look, this is 235 00:12:49,559 --> 00:12:53,240 Speaker 2: too fast moving, this is too complicated, and the legislators 236 00:12:53,280 --> 00:12:57,200 Speaker 2: just don't have the technical expertise to regulate. But I 237 00:12:57,240 --> 00:13:00,840 Speaker 2: would rather say that this is not just a part technology. 238 00:13:01,280 --> 00:13:06,040 Speaker 2: It is also about how technology implicates fundamental rights, how 239 00:13:06,080 --> 00:13:10,439 Speaker 2: it implicates democracy, And I wouldn't say that the facebooks 240 00:13:10,480 --> 00:13:13,720 Speaker 2: of the world would have any expertise in being in 241 00:13:13,800 --> 00:13:17,920 Speaker 2: charge of democracy and fundamental rights. So that alone, I 242 00:13:17,960 --> 00:13:22,160 Speaker 2: think justifies that democratic governance have the seat at the 243 00:13:22,200 --> 00:13:25,560 Speaker 2: table when it comes to steering the AI future of 244 00:13:25,679 --> 00:13:26,480 Speaker 2: the societies. 245 00:13:28,320 --> 00:13:32,160 Speaker 10: I think that you have all hurt and probably agree 246 00:13:32,559 --> 00:13:36,440 Speaker 10: that AI is too important not to regulate, and it's 247 00:13:36,440 --> 00:13:42,040 Speaker 10: too important to badly regulate. A good regulation that we 248 00:13:42,200 --> 00:13:45,120 Speaker 10: all agree on as soon as possible must be a 249 00:13:45,160 --> 00:13:46,079 Speaker 10: common objective. 250 00:13:46,160 --> 00:13:49,720 Speaker 4: What you just heard was Marguerite Vestiger, the EU Commission's 251 00:13:49,760 --> 00:13:55,360 Speaker 4: executive Vice President, addressing the European Parliament last month. So 252 00:13:55,440 --> 00:13:58,320 Speaker 4: would you describe the legislation as sort of the best 253 00:13:58,400 --> 00:14:01,040 Speaker 4: that it can be at this point in time or 254 00:14:01,080 --> 00:14:03,360 Speaker 4: when you look at it given your experience with sort 255 00:14:03,400 --> 00:14:07,000 Speaker 4: of regulation and the challenges as a whole, do you 256 00:14:07,040 --> 00:14:09,760 Speaker 4: see any holes in the legislation? Do you see things 257 00:14:09,840 --> 00:14:13,160 Speaker 4: that lawmakers have missed in this process? So what would 258 00:14:13,200 --> 00:14:15,679 Speaker 4: be your advice to them if you could look at 259 00:14:15,679 --> 00:14:17,439 Speaker 4: this legislation and talk to them about it? 260 00:14:18,080 --> 00:14:21,320 Speaker 2: So I think it would be probably fair to say 261 00:14:21,440 --> 00:14:24,400 Speaker 2: that this is not going to be a perfect legislation 262 00:14:24,720 --> 00:14:28,560 Speaker 2: and that sometimes then invites the question whether no legislation 263 00:14:28,720 --> 00:14:32,400 Speaker 2: is better than imperfect legislation, And there I'm willing to 264 00:14:32,440 --> 00:14:35,640 Speaker 2: say that even if we don't fully get to the 265 00:14:35,760 --> 00:14:38,960 Speaker 2: kind of regulation that best serves the development of AI 266 00:14:39,440 --> 00:14:43,160 Speaker 2: over the future years, it is still important to get 267 00:14:43,400 --> 00:14:47,520 Speaker 2: basic guardrails in place. So yes, in many ways, I 268 00:14:47,560 --> 00:14:50,880 Speaker 2: think the legislation is still somewhat vague, so that I 269 00:14:50,880 --> 00:14:52,960 Speaker 2: think suggests this is the best that we can do 270 00:14:53,400 --> 00:14:57,320 Speaker 2: at this point. You cannot expect the lawmakers to become 271 00:14:57,400 --> 00:15:00,240 Speaker 2: too specific in prescribing how you dive a lot of 272 00:15:00,320 --> 00:15:03,920 Speaker 2: these technologies. But at the same time, they do have 273 00:15:04,040 --> 00:15:08,200 Speaker 2: the need to put in place those transparency applications so 274 00:15:08,280 --> 00:15:11,600 Speaker 2: that we can still have democratic oversight and a conversation 275 00:15:11,760 --> 00:15:15,280 Speaker 2: that we open the black box behind these technologies and 276 00:15:15,400 --> 00:15:17,040 Speaker 2: gain a better understanding. 277 00:15:17,600 --> 00:15:20,720 Speaker 4: The EU approach is very different from the US approach, 278 00:15:20,720 --> 00:15:23,640 Speaker 4: which is very hands off, basically sort of looking towards 279 00:15:23,720 --> 00:15:26,520 Speaker 4: voluntary regulation if there is such a thing. 280 00:15:27,200 --> 00:15:30,160 Speaker 11: AI can help deal with some very difficult challenges like 281 00:15:30,240 --> 00:15:33,480 Speaker 11: disease and climate change, but also have to address the 282 00:15:33,480 --> 00:15:36,480 Speaker 11: potential risk to our society, to our economy, to our 283 00:15:36,560 --> 00:15:39,920 Speaker 11: national security. Tech companies have a responsibility in my view, 284 00:15:39,960 --> 00:15:43,640 Speaker 11: to make sure the products are safe before making them public. 285 00:15:44,240 --> 00:15:46,960 Speaker 4: I'm interested in your view on that. And b even 286 00:15:47,000 --> 00:15:49,720 Speaker 4: if the US decided they wanted to tackle this, is 287 00:15:49,760 --> 00:15:52,320 Speaker 4: it just impossible during the US election season. 288 00:15:52,920 --> 00:15:54,840 Speaker 9: So the US has for. 289 00:15:54,880 --> 00:16:00,240 Speaker 2: Quite a while been reluctant or unable to legislate soared 290 00:16:00,280 --> 00:16:03,440 Speaker 2: to the European, Americans are still much more convinced that 291 00:16:03,480 --> 00:16:06,200 Speaker 2: there is faith in the operation of the markets and 292 00:16:06,400 --> 00:16:09,800 Speaker 2: greater faith in the tech companies self regulation, and at 293 00:16:09,800 --> 00:16:13,320 Speaker 2: the same time greater hesitancy that the governments can improve 294 00:16:13,440 --> 00:16:17,120 Speaker 2: outcomes by stepping in. So that's why I've often described 295 00:16:17,160 --> 00:16:20,200 Speaker 2: the American way of regulating more of a market based 296 00:16:20,240 --> 00:16:24,200 Speaker 2: approach as opposed to the European rights based approach, and 297 00:16:24,240 --> 00:16:24,800 Speaker 2: I think there are. 298 00:16:24,720 --> 00:16:25,800 Speaker 9: A few reasons for that. 299 00:16:26,080 --> 00:16:31,120 Speaker 2: So the Americans are still very focused on safeguarding the 300 00:16:31,200 --> 00:16:36,360 Speaker 2: conditions for a thriving tech ecosystem that is ideal for innovation. 301 00:16:36,800 --> 00:16:40,160 Speaker 2: They are also regulating in the shadow of US China 302 00:16:40,200 --> 00:16:44,240 Speaker 2: tech war, so Americans are very concerned that the US 303 00:16:44,360 --> 00:16:48,080 Speaker 2: will retain its technological supremacy and it cannot be left 304 00:16:48,120 --> 00:16:52,240 Speaker 2: behind in this ongoing tech race. There's also tremendous lobbying 305 00:16:52,320 --> 00:16:55,800 Speaker 2: by the tech companies that partially explain why the US 306 00:16:56,040 --> 00:17:00,160 Speaker 2: Congress has not been able to generate any meaningful legisla 307 00:17:00,760 --> 00:17:03,560 Speaker 2: even though the public opinion in the US has been 308 00:17:03,600 --> 00:17:07,760 Speaker 2: shifting to be more skeptical of tech companies self regulation, 309 00:17:08,400 --> 00:17:12,520 Speaker 2: and it's just generally that the political process is rather dysfunctional, 310 00:17:12,640 --> 00:17:18,080 Speaker 2: we see very little meaningful legislation emerging from Congress. Still today, 311 00:17:18,160 --> 00:17:22,080 Speaker 2: we are seeing something like a voluntary guidance coming from 312 00:17:22,160 --> 00:17:26,479 Speaker 2: the White House. There's a blueprint for AI Bill of Rights, 313 00:17:26,560 --> 00:17:32,320 Speaker 2: but ultimately those voluntary regulations, they leave the decisions in 314 00:17:32,359 --> 00:17:33,760 Speaker 2: the hands of the tech companies. 315 00:17:35,200 --> 00:17:37,680 Speaker 4: In terms of talking about the mitigation part of it, 316 00:17:37,760 --> 00:17:41,960 Speaker 4: do you see a potential to actually use AI to 317 00:17:42,080 --> 00:17:45,560 Speaker 4: regulate AI? Are there any aspects of AI that can 318 00:17:45,600 --> 00:17:48,640 Speaker 4: be drawn out that are more positive, that are more productive, 319 00:17:48,680 --> 00:17:51,720 Speaker 4: that can be used in the application of systems to 320 00:17:51,800 --> 00:17:52,679 Speaker 4: regulate AI. 321 00:17:53,520 --> 00:17:58,159 Speaker 2: I think it's almost inevitable. Ultimately, if the AAI really 322 00:17:58,320 --> 00:18:01,240 Speaker 2: is as powerful of a tool, for instance, to be 323 00:18:01,359 --> 00:18:04,520 Speaker 2: used by criminals, we need to then make sure that 324 00:18:04,600 --> 00:18:07,560 Speaker 2: we also have the good actors to deploy the AI 325 00:18:07,640 --> 00:18:10,200 Speaker 2: and develop AI in a way that we can, for instance, 326 00:18:10,240 --> 00:18:14,520 Speaker 2: more easily detect the wrongdoing, we can detect fraud, and 327 00:18:14,560 --> 00:18:18,480 Speaker 2: if we think about using AI to spread fraudulent information 328 00:18:18,760 --> 00:18:22,920 Speaker 2: misinformation that we also have the power of that technology 329 00:18:22,920 --> 00:18:25,959 Speaker 2: that is deployed in a way that allows us to 330 00:18:26,000 --> 00:18:30,119 Speaker 2: fight those downsides and mitigate that kind of harmful activity. 331 00:18:30,359 --> 00:18:33,760 Speaker 2: So absolutely, I think there is a great need and 332 00:18:33,840 --> 00:18:36,879 Speaker 2: I would like to think a great incentive by many 333 00:18:36,920 --> 00:18:40,520 Speaker 2: developers of these technologies to make sure that the technology 334 00:18:40,600 --> 00:18:42,959 Speaker 2: is deployed in a way that can be used as 335 00:18:43,000 --> 00:18:44,400 Speaker 2: a powerful weapon for good. 336 00:18:44,920 --> 00:18:46,679 Speaker 4: As we come back to the idea where the EU 337 00:18:47,280 --> 00:18:49,800 Speaker 4: is that, could you just walk us through where the 338 00:18:49,840 --> 00:18:52,760 Speaker 4: act is at now and what the processes are to 339 00:18:52,800 --> 00:18:55,040 Speaker 4: bring it fully into law because it has to be 340 00:18:55,119 --> 00:18:57,399 Speaker 4: discussed with the EU member states and there could be 341 00:18:57,480 --> 00:18:59,160 Speaker 4: some wrinkles, no doubt there. 342 00:18:59,480 --> 00:19:01,200 Speaker 9: What are the next steps in this process? 343 00:19:01,920 --> 00:19:04,920 Speaker 2: So we two years into this process rather so that's 344 00:19:04,960 --> 00:19:08,919 Speaker 2: when the proposal was first put forward by the European Commission, 345 00:19:09,400 --> 00:19:12,560 Speaker 2: and right now we have the proposal that has gone 346 00:19:12,600 --> 00:19:16,080 Speaker 2: through the legislative process in the Council that represents the 347 00:19:16,160 --> 00:19:19,680 Speaker 2: voices of the Member States and in the Parliament that 348 00:19:19,800 --> 00:19:23,840 Speaker 2: then represents the European citizens. But now we have this 349 00:19:24,119 --> 00:19:27,199 Speaker 2: last phase of the legislative process where we need to 350 00:19:27,280 --> 00:19:31,199 Speaker 2: reconcile the differences between the Parliament and the Council, and 351 00:19:31,320 --> 00:19:35,080 Speaker 2: the Commission is there as a broker in these conversations, 352 00:19:35,359 --> 00:19:37,560 Speaker 2: So there are some differences. I think it's fair to 353 00:19:37,600 --> 00:19:42,000 Speaker 2: say that the Parliament has been more ambitious in imposing 354 00:19:42,160 --> 00:19:46,000 Speaker 2: additional applications that we're not part of the version that 355 00:19:46,160 --> 00:19:49,840 Speaker 2: went through the Council. And now this fall is the 356 00:19:49,920 --> 00:19:53,240 Speaker 2: time when we are trying to find a consensus between 357 00:19:53,280 --> 00:19:57,679 Speaker 2: the two legislators, and because of the looming European elections 358 00:19:57,720 --> 00:20:00,520 Speaker 2: next summer, there is a timing issue. 359 00:20:00,720 --> 00:20:03,320 Speaker 9: So if we fail to reach. 360 00:20:03,119 --> 00:20:08,200 Speaker 2: A consensus and pass the AI Act through this final stage, 361 00:20:08,520 --> 00:20:11,399 Speaker 2: the fear is that we go to the next elections 362 00:20:11,480 --> 00:20:13,919 Speaker 2: and then we have a new slate of maps with 363 00:20:14,040 --> 00:20:16,800 Speaker 2: their own priorities, and we need to go back to 364 00:20:16,840 --> 00:20:20,720 Speaker 2: the drawing board. So there is a great need now 365 00:20:20,800 --> 00:20:25,719 Speaker 2: for Spain, as the President of the European Council, Douta 366 00:20:25,920 --> 00:20:28,399 Speaker 2: try to proger the compromise and make sure that we 367 00:20:28,480 --> 00:20:30,520 Speaker 2: can get the law out. Still in the course of 368 00:20:30,520 --> 00:20:32,360 Speaker 2: this fall, of. 369 00:20:32,320 --> 00:20:34,440 Speaker 4: Course, I need to ask you about the Brussels effect 370 00:20:34,440 --> 00:20:36,720 Speaker 4: which is a phrase that many of our listeners would 371 00:20:36,720 --> 00:20:38,159 Speaker 4: have heard of but may not know. The EU in 372 00:20:38,240 --> 00:20:41,320 Speaker 4: fact coined this phrase, which is to describe the way 373 00:20:41,359 --> 00:20:44,760 Speaker 4: that essentially, if I've got this RIEU regulation becomes the 374 00:20:44,840 --> 00:20:48,240 Speaker 4: standard everywhere because companies say, well, we need to standardize 375 00:20:48,240 --> 00:20:52,080 Speaker 4: our operations globally, so even though it's only an EU regulation, 376 00:20:52,200 --> 00:20:55,400 Speaker 4: it can become the way they operate, including. 377 00:20:55,240 --> 00:20:55,919 Speaker 9: In the US. 378 00:20:56,520 --> 00:20:59,119 Speaker 4: Would you expect this to be the same in this case? 379 00:21:00,400 --> 00:21:03,679 Speaker 2: I believe that there is very likely to be a 380 00:21:03,800 --> 00:21:07,960 Speaker 2: Brussels effect, at least on some applications of AI. So, 381 00:21:08,040 --> 00:21:10,480 Speaker 2: first of all, AI is a multifaceted thing and there 382 00:21:10,480 --> 00:21:14,600 Speaker 2: are many forms and types of AI, so I wouldn't 383 00:21:14,920 --> 00:21:18,639 Speaker 2: think that we will see every AI system to be 384 00:21:18,800 --> 00:21:22,080 Speaker 2: following the European regulations around the world. But there are 385 00:21:22,119 --> 00:21:25,879 Speaker 2: a couple of features of how AI works that lends 386 00:21:25,920 --> 00:21:29,080 Speaker 2: itself really well to the Brussels effect. So one is 387 00:21:29,080 --> 00:21:32,560 Speaker 2: that you need a lot of data to develop empower 388 00:21:32,680 --> 00:21:35,800 Speaker 2: these models. So the more data you have normally the 389 00:21:35,880 --> 00:21:40,280 Speaker 2: better AI models you have. And if you are willing 390 00:21:40,359 --> 00:21:44,359 Speaker 2: to use the European data to train these models, you 391 00:21:44,560 --> 00:21:48,320 Speaker 2: are bound by European AI Act and then if you 392 00:21:48,440 --> 00:21:52,160 Speaker 2: want to offer that same AI system in another market 393 00:21:52,480 --> 00:21:55,840 Speaker 2: and use those models that were trained using the European data, 394 00:21:55,960 --> 00:21:58,600 Speaker 2: you do need to continue to be bounded by the 395 00:21:59,359 --> 00:22:03,640 Speaker 2: AI Act. So if you wanted to evade the European 396 00:22:03,680 --> 00:22:07,800 Speaker 2: regulations and for instance, offer them the same AI products 397 00:22:07,840 --> 00:22:10,520 Speaker 2: somewhere in the United States, you would need to retrain 398 00:22:10,640 --> 00:22:14,160 Speaker 2: the model and carve out all the European data from 399 00:22:14,160 --> 00:22:18,200 Speaker 2: that model. So that is often unappealing for companies. 400 00:22:19,400 --> 00:22:22,320 Speaker 4: When we return more from Bloomberg's Gillian Deutsch about the 401 00:22:22,320 --> 00:22:34,760 Speaker 4: future of AI regulation, Jillian, let's talk a bit about 402 00:22:34,760 --> 00:22:37,159 Speaker 4: the role of big tech in their lobbying efforts that 403 00:22:37,200 --> 00:22:40,560 Speaker 4: have gone on so far. Obviously, tech companies themselves, some 404 00:22:40,560 --> 00:22:43,320 Speaker 4: of them expressed concern the CEOs about the pace of 405 00:22:43,400 --> 00:22:45,920 Speaker 4: AI and the risk from AI, but also they want 406 00:22:45,920 --> 00:22:47,760 Speaker 4: to harness it and make money from it. 407 00:22:47,840 --> 00:22:49,720 Speaker 9: So what have they been lobbying the EU for. 408 00:22:50,560 --> 00:22:53,640 Speaker 6: It's interesting because the EU proposed AIA Act two years 409 00:22:53,680 --> 00:22:56,600 Speaker 6: ago and companies have been lobbying like crazy in the 410 00:22:56,640 --> 00:22:59,440 Speaker 6: interim and actually they elected to the Commission's first proposal 411 00:22:59,560 --> 00:23:03,080 Speaker 6: because for the most part it left generative AI untouched. 412 00:23:03,440 --> 00:23:06,040 Speaker 6: It was mostly if this technology was used in high 413 00:23:06,119 --> 00:23:09,480 Speaker 6: risk circumstances, then it would be regulated. So tech companies 414 00:23:09,520 --> 00:23:11,879 Speaker 6: were mostly pretty happy with that initial idea. It was 415 00:23:11,920 --> 00:23:16,080 Speaker 6: when you countries and lawmakers began adding more controls that 416 00:23:16,119 --> 00:23:18,800 Speaker 6: they started to ring alarm bells and freak out. It's 417 00:23:18,880 --> 00:23:21,480 Speaker 6: kind of difficult to summarize exactly what companies all want, 418 00:23:21,600 --> 00:23:23,240 Speaker 6: but if I were to boil it down to three points, 419 00:23:23,280 --> 00:23:26,919 Speaker 6: it's basically one trust us week as companies have already 420 00:23:26,920 --> 00:23:29,800 Speaker 6: put in our own controls. We have not launched certain 421 00:23:29,800 --> 00:23:33,080 Speaker 6: products as well because of possible misuse or problems we've 422 00:23:33,080 --> 00:23:35,800 Speaker 6: seen arise. Two, they say it's not really up to 423 00:23:35,880 --> 00:23:39,200 Speaker 6: us as developers to see how this technology is used. 424 00:23:39,480 --> 00:23:40,479 Speaker 7: It's up to the users. 425 00:23:40,520 --> 00:23:42,800 Speaker 6: It's up to these companies that purchase that AI. They're 426 00:23:42,800 --> 00:23:44,800 Speaker 6: the ones who determine how it's actually used. So it's 427 00:23:44,920 --> 00:23:48,240 Speaker 6: really up to those companies and those users to bear the. 428 00:23:48,160 --> 00:23:50,560 Speaker 7: Cost and the burden of compliance of regulation. 429 00:23:50,840 --> 00:23:52,880 Speaker 6: And Third, they really want to stick with this risk 430 00:23:52,920 --> 00:23:55,440 Speaker 6: based approach, which, to remind you, that means very few 431 00:23:55,480 --> 00:23:56,879 Speaker 6: controls on genertive AI. 432 00:23:57,440 --> 00:24:00,800 Speaker 4: And how will this fit against individual country regulation? I mean, 433 00:24:00,840 --> 00:24:04,480 Speaker 4: obviously nations sort of do agree that Brussels has the 434 00:24:04,560 --> 00:24:07,120 Speaker 4: say on some of these broad issues, but equally each 435 00:24:07,160 --> 00:24:09,160 Speaker 4: country is still going to try and do their own thing. 436 00:24:09,640 --> 00:24:13,399 Speaker 6: To some extent, I think individual countries are really going 437 00:24:13,440 --> 00:24:16,240 Speaker 6: to be tripping over each other to become the AI 438 00:24:16,320 --> 00:24:19,879 Speaker 6: hub in the EU because the EU has a massive 439 00:24:19,920 --> 00:24:20,760 Speaker 6: chip on his shoulder. 440 00:24:21,040 --> 00:24:22,800 Speaker 7: They've missed out on many waves of. 441 00:24:22,720 --> 00:24:26,639 Speaker 6: Technology, most notably the social media wave. They want to 442 00:24:26,640 --> 00:24:30,240 Speaker 6: have some kind of company that could someday really rival 443 00:24:30,240 --> 00:24:32,919 Speaker 6: the lakes of Google, the lakes of Microsoft. And so 444 00:24:32,960 --> 00:24:35,960 Speaker 6: we will definitely see countries like France, Spain trying to 445 00:24:36,000 --> 00:24:39,080 Speaker 6: attract those startups, trying to build them up, help them 446 00:24:39,080 --> 00:24:41,320 Speaker 6: scale up, and not lose them to the US like 447 00:24:41,359 --> 00:24:43,960 Speaker 6: they've seen in previous startups. So we definitely will see 448 00:24:43,960 --> 00:24:47,560 Speaker 6: countries trying to become the most hospitable hub for AI. 449 00:24:48,480 --> 00:24:50,520 Speaker 4: And we've talked a lot, I guess about the risks 450 00:24:50,600 --> 00:24:53,840 Speaker 4: about the full lad about slightly dystopian future that people 451 00:24:53,920 --> 00:24:57,359 Speaker 4: warn might come with AI. So let's maybe talk just 452 00:24:57,400 --> 00:25:00,879 Speaker 4: briefly about the upside. Also when we look at AI 453 00:25:01,359 --> 00:25:05,080 Speaker 4: and the possibilities across business, governments and so on, what 454 00:25:05,119 --> 00:25:08,840 Speaker 4: does the EU sees the potential of AI that is. 455 00:25:08,840 --> 00:25:10,880 Speaker 6: Such a great question because so much of the conversation 456 00:25:10,920 --> 00:25:13,960 Speaker 6: here in Brussels is about the possible harms and big 457 00:25:14,000 --> 00:25:17,639 Speaker 6: tech companies trying to steal the entire market in Europe 458 00:25:17,720 --> 00:25:18,840 Speaker 6: never being able to compete. 459 00:25:19,080 --> 00:25:20,480 Speaker 7: I think lawmakers do. 460 00:25:20,520 --> 00:25:26,199 Speaker 6: Realize that there is serious upside to productivity to the 461 00:25:26,240 --> 00:25:28,560 Speaker 6: healthcare industry and they want to make sure that. 462 00:25:28,520 --> 00:25:30,240 Speaker 7: They also can capitalize on that as well. 463 00:25:30,440 --> 00:25:33,280 Speaker 6: Part of the A Act actually is this thing called sandboxes, 464 00:25:33,359 --> 00:25:37,800 Speaker 6: which is basically allowing smaller companies to operate with few 465 00:25:37,880 --> 00:25:40,000 Speaker 6: restrictions so that they can kind of test out their 466 00:25:40,040 --> 00:25:43,359 Speaker 6: technology and not be constrained by so much regulation. And 467 00:25:43,400 --> 00:25:46,480 Speaker 6: so this at least is one opportunity for EU politicians 468 00:25:46,520 --> 00:25:48,639 Speaker 6: to say, look, we're not just focused in the regulation, 469 00:25:49,040 --> 00:25:50,960 Speaker 6: we also care about the innovation here as well. 470 00:25:51,760 --> 00:25:53,840 Speaker 4: If you had to make a bet, what would you 471 00:25:53,880 --> 00:25:55,640 Speaker 4: say would be in that final bill. 472 00:25:55,840 --> 00:25:58,680 Speaker 6: There's no question to me generative AI will be included 473 00:25:58,880 --> 00:26:02,480 Speaker 6: in some way. I do think that most people that 474 00:26:02,520 --> 00:26:04,119 Speaker 6: are not in the Parliament I think the Parliament. 475 00:26:03,880 --> 00:26:04,360 Speaker 7: Went too far. 476 00:26:04,800 --> 00:26:08,280 Speaker 6: So I think some of these concerns or from companies 477 00:26:08,680 --> 00:26:11,680 Speaker 6: that they will not be able to quantify the impact 478 00:26:11,680 --> 00:26:13,880 Speaker 6: on the rule of law or the impact on the environment. 479 00:26:14,400 --> 00:26:17,080 Speaker 6: Those risk classments will probably be taken out, but we 480 00:26:17,080 --> 00:26:19,080 Speaker 6: will definitely will see some controls of Genervia, There's no 481 00:26:19,160 --> 00:26:19,840 Speaker 6: question about that. 482 00:26:20,920 --> 00:26:22,480 Speaker 9: Julian, thank you very much for your time. 483 00:26:22,680 --> 00:26:25,560 Speaker 4: Thank you so much for as Thanks for listening to 484 00:26:25,640 --> 00:26:27,840 Speaker 4: us here at The Big Take. It's a daily podcast 485 00:26:27,840 --> 00:26:32,160 Speaker 4: from Bloomberg and iHeartRadio. For more shows from iHeartRadio, visit 486 00:26:32,240 --> 00:26:36,800 Speaker 4: the iHeartRadio app, Apple Podcasts, or wherever you listen, and 487 00:26:36,840 --> 00:26:39,480 Speaker 4: we'd love to hear from you. Email us with questions 488 00:26:39,520 --> 00:26:43,040 Speaker 4: or comments to Big Take at Bloomberg dot net. The 489 00:26:43,080 --> 00:26:46,720 Speaker 4: supervising producer of The Big Take is Vicky Viergelina. Our 490 00:26:46,760 --> 00:26:51,440 Speaker 4: senior producer is Catherine Fink. Federica Romangniello is our producer. 491 00:26:52,200 --> 00:26:57,000 Speaker 4: Our associate producer is zenab Zidiki. Hilde Garcia is our engineer. 492 00:26:57,320 --> 00:27:02,200 Speaker 4: Original music by Leo Sidrin. I'm Rosland Matheson in for Wescousover. 493 00:27:02,640 --> 00:27:04,480 Speaker 4: We'll be back tomorrow with another Big Take.