Startup ‘Fake It’ Culture Isn’t an Excuse for Fraud, Says SEC

[00:00:03] [SPEAKER_00]: Exchanges, the Goldman Sachs podcast featuring exchanges on rates, inflation, and U.S. recession risk.

[00:00:12] [SPEAKER_00]: Exchanges on the market impact of AI.

[00:00:15] [SPEAKER_00]: For the sharpest analysis on forces driving the markets and the economy,

[00:00:20] [SPEAKER_00]: count on exchanges between the leading minds at Goldman Sachs.

[00:00:24] [SPEAKER_00]: New episodes every week.

[00:00:25] [SPEAKER_00]: Listen now.

[00:00:33] [SPEAKER_03]: Welcome to Tech News Briefing.

[00:00:35] [SPEAKER_03]: It's Friday, October 4th.

[00:00:38] [SPEAKER_03]: I'm Zoe Thomas for The Wall Street Journal.

[00:00:41] [SPEAKER_03]: Off-the-shelf large-language artificial intelligence models like ChatGPT and Claude know a lot,

[00:00:48] [SPEAKER_03]: but maybe not what companies need them to know.

[00:00:51] [SPEAKER_03]: That's forcing companies to augment today's general models with industry or business-specific

[00:00:57] [SPEAKER_03]: information to make them useful.

[00:00:59] [SPEAKER_03]: We'll tell you more.

[00:01:01] [SPEAKER_03]: And then, Silicon Valley's startup culture loves a good motto.

[00:01:07] [SPEAKER_03]: Move fast and break things.

[00:01:08] [SPEAKER_03]: Ask forgiveness, not permission.

[00:01:11] [SPEAKER_03]: Startups that win, keep winning.

[00:01:13] [SPEAKER_03]: But financial regulators are sending a message that fake it till you make it isn't an excuse

[00:01:19] [SPEAKER_03]: for fraud.

[00:01:20] [SPEAKER_03]: Our reporter Mark Vardabedian is going to tell us about the efforts to crack down on startups

[00:01:24] [SPEAKER_03]: misleading or defrauding investors.

[00:01:27] [SPEAKER_03]: But first, generative AI's foundation models can be trained on vast troves of online data,

[00:01:37] [SPEAKER_03]: but still lack deep-specific knowledge even on mainstream topics.

[00:01:42] [SPEAKER_03]: That means companies have to take additional steps to make these models more useful.

[00:01:46] [SPEAKER_03]: Here to tell us more is WSJ CIO Journal reporter Isabel Busquette.

[00:01:52] [SPEAKER_03]: So, Isabel, you write in your story that today's AI models are about as useful out of the box as a new employee

[00:02:00] [SPEAKER_03]: entering orientation.

[00:02:01] [SPEAKER_03]: Why is that?

[00:02:03] [SPEAKER_02]: The models know and understand the information they've been trained on.

[00:02:09] [SPEAKER_02]: And they've been trained on a ton of information, all this stuff that's publicly available on the internet they've been trained on.

[00:02:18] [SPEAKER_02]: So, in theory, they should know a lot.

[00:02:20] [SPEAKER_02]: But when you're a business and you're in a specific industry or any kind of organization,

[00:02:26] [SPEAKER_02]: and so one of the ones I talked to for this story was the PGA Tour, the golf organization.

[00:02:32] [SPEAKER_02]: And they found that, you know, when they went on to chat GPT, it couldn't really answer specific questions about golf.

[00:02:39] [SPEAKER_02]: Like, it couldn't give them an accurate answer on how many PGA Tours Tiger Woods has won.

[00:02:45] [SPEAKER_02]: It couldn't understand the difference between a tour win and a majors win.

[00:02:49] [SPEAKER_02]: And so, that's the kind of specific detailed information that these models don't necessarily have out of the box.

[00:02:58] [SPEAKER_02]: And that's why it's not so straightforward to just set it up in a business use case.

[00:03:03] [SPEAKER_02]: And the companies are finding they actually have to do a lot of work on their end to customize these models

[00:03:09] [SPEAKER_02]: and augment them with all their own data before they can really put them to work.

[00:03:13] [SPEAKER_03]: What are some of the challenges of doing that, of getting them to have the right data that makes them work for a specific company?

[00:03:20] [SPEAKER_02]: You know, if you're a company and you want to augment a model based on your own data,

[00:03:25] [SPEAKER_02]: the first obstacle is that you have a good handle on your data.

[00:03:29] [SPEAKER_02]: And the truth is that most companies don't.

[00:03:32] [SPEAKER_02]: Their data is kind of a mess.

[00:03:34] [SPEAKER_02]: They don't know where it is.

[00:03:35] [SPEAKER_02]: They have conflicting versions of the same data.

[00:03:37] [SPEAKER_02]: This is an issue they've struggled with for years and years and years.

[00:03:40] [SPEAKER_02]: But they really have to kind of sort that out if they want to begin the process of augmenting these models with their data.

[00:03:46] [SPEAKER_02]: They're then faced with a series of options of how to go about this.

[00:03:51] [SPEAKER_02]: And obviously, the more you put into this, the more accurate the output is going to be.

[00:03:57] [SPEAKER_02]: But also, you know, the more time and money you're spending.

[00:04:00] [SPEAKER_02]: So, companies have to decide how much do we want to put it into this?

[00:04:05] [SPEAKER_02]: How accurate do we really need this model to be?

[00:04:07] [SPEAKER_02]: Some of it depends on how high stakes the use cases are.

[00:04:12] [SPEAKER_02]: So, the agriculture unit at Bayer is experimenting with some of this stuff.

[00:04:17] [SPEAKER_02]: And they have one use case where they're using AI to help answer questions for engineers as they get onboarded.

[00:04:25] [SPEAKER_02]: And this is a use case where it's low stakes.

[00:04:29] [SPEAKER_02]: It's internal employees.

[00:04:30] [SPEAKER_02]: They'll figure out the right answer eventually.

[00:04:33] [SPEAKER_02]: Whereas they're also thinking about models that are going to maybe be giving advice to farmers about how exactly to tend their crops this season.

[00:04:43] [SPEAKER_02]: And there, the stakes are extremely high.

[00:04:45] [SPEAKER_02]: If something bad happens with the crops, that's, you know, a ton of money and basically the entire business.

[00:04:50] [SPEAKER_02]: All right.

[00:04:51] [SPEAKER_03]: That was our reporter, Isabel Busquette.

[00:04:53] [SPEAKER_03]: Coming up, startup culture has long encouraged setting lofty or even unrealistic growth projections.

[00:05:01] [SPEAKER_03]: Regulators say that isn't an excuse for fraud.

[00:05:04] [SPEAKER_03]: We'll have more on that after the break.

[00:05:18] [SPEAKER_03]: Regulators have a message for startups.

[00:05:20] [SPEAKER_03]: Fake it till you bake it can be fraud.

[00:05:23] [SPEAKER_03]: A series of recent enforcement actions by the Securities and Exchange Commission is driving home that message.

[00:05:30] [SPEAKER_03]: Last month, the regulator charged a startup and some of the founders or executives of two others with misleading or defrauding investors.

[00:05:40] [SPEAKER_03]: Here to tell us more is Mark Vardabedian, a reporter with the WSJ Pro Venture Capital team.

[00:05:45] [SPEAKER_03]: So, Mark, startup culture has long encouraged setting lofty goals or even unrealistic growth projections to attract investors.

[00:05:53] [SPEAKER_03]: So why is the SEC cracking down on this now?

[00:05:57] [SPEAKER_01]: This is coming in the wake of some very high-profile startup collapses that have happened in recent years.

[00:06:04] [SPEAKER_01]: Thinking about Theranos and FTX in particular.

[00:06:08] [SPEAKER_01]: You know, those were cases where startups had very high-profile investors and grand ambitions that went awry and very obviously went beyond the lofty ambitions of fake it till you make it and entered into the realm of fraud.

[00:06:24] [SPEAKER_03]: What specifically have we heard from folks at the SEC?

[00:06:28] [SPEAKER_01]: In announcing the charges, the SEC was pretty clear about their messaging.

[00:06:32] [SPEAKER_01]: The director of the SEC's San Francisco office said in the statement of one of the charges, pretty much amounting to startup founders cannot fake it until they make it by falsifying revenue metrics shared with investors.

[00:06:44] [SPEAKER_03]: You mentioned some of the well-publicized collapses like blood testing company Theranos and there's also crypto company FTX.

[00:06:52] [SPEAKER_03]: There are these other charges that have been brought against startups.

[00:06:56] [SPEAKER_03]: Tell us about business automation startup scale.

[00:06:59] [SPEAKER_01]: Scale is a company that the SEC really targeted here.

[00:07:04] [SPEAKER_01]: They allege that the co-founder and chief executive Baba Nadeh Mepali managed to raise more than $30 million from investors between 2021 and 2022 by claiming it had as much as $7 million in annual recurring revenue when that actual figure was no more than $170,000.

[00:07:26] [SPEAKER_03]: I mean, that's a huge gap.

[00:07:28] [SPEAKER_03]: Gap, who were some of Scale's investors?

[00:07:30] [SPEAKER_01]: Yeah, Scale had a couple investors including RTP Global, kind of your venture firm focused on early and mid-stage companies.

[00:07:38] [SPEAKER_01]: They declined to speak with us for this article and so did Nada Pilami.

[00:07:43] [SPEAKER_03]: Let's talk about another company, Medley Health.

[00:07:46] [SPEAKER_03]: It had some really high-profile investors.

[00:07:49] [SPEAKER_03]: What charges has the SEC brought against Medley executives?

[00:07:52] [SPEAKER_01]: Medley ran pharmacies that people could go to for prescriptions and offered a digital system that could help them manage prescriptions and get pills delivered and other health care-related services.

[00:08:07] [SPEAKER_01]: So the SEC alleged that three executives defrauded investors during capital raising efforts, and that ultimately landed the now defunct digital health startup over $170 million.

[00:08:19] [SPEAKER_01]: The most glaring example of fraud in the charges were from at least February 2021 through August 2022.

[00:08:27] [SPEAKER_01]: Two of the executives provided this phony financial information to prospective investors.

[00:08:33] [SPEAKER_01]: And part of that was based on totally fake prescriptions that a third executive had logged into the system.

[00:08:41] [SPEAKER_01]: So very clear instance of fraud in that case, according to the SEC.

[00:08:45] [SPEAKER_01]: And Medley had raised capital from venture firm Graycroft, a well-known firm in the Valley.

[00:08:52] [SPEAKER_03]: We should note Graycroft didn't respond to a request for comment, and Medley's co-founder and former CEO, one of the executives named in the SEC's case, didn't respond to a request for comment either.

[00:09:05] [SPEAKER_03]: Mark, there's a third company that's facing charges from the SEC, Zimergen.

[00:09:11] [SPEAKER_03]: Tell us what the company does and what the SEC is alleging against them.

[00:09:15] [SPEAKER_01]: Zimergen is an Emeryville-based biotechnology company, and the SEC's charges against Zimergen is that it misled IPO investors about its market potential, revenue prospects, and also its customers.

[00:09:29] [SPEAKER_01]: And ultimately, the company raised about $530 million through its IPO before it filed for bankruptcy.

[00:09:36] [SPEAKER_01]: And the SEC alleged that Zimergen, similar to the other companies, misled investors and fudged some of their revenue figures and prospects related to customers.

[00:09:50] [SPEAKER_01]: So the common thread in all of these is misleading investors and making up business metrics to attract investors.

[00:09:58] [SPEAKER_03]: In the end, Zimergen reached a deal with the SEC to cease and desist orders and pay this civil penalty without admitting or denying what the SEC found.

[00:10:11] [SPEAKER_03]: I'm curious, though, how much of this responsibility to check that companies are doing what they say they're going to do and bringing in the revenue that they say they're bringing in, how much of that should be or is on investors?

[00:10:23] [SPEAKER_01]: Investors will tell you startup fraud in Silicon Valley is extremely rare.

[00:10:28] [SPEAKER_01]: They also acknowledge that in this world of providing lofty or ambitious outlooks, they've perhaps been lulled into a false sense of security and that they don't always dig as deep into some of the startups as they should.

[00:10:41] [SPEAKER_01]: And going back to Theranos and FTX, those collapses have really spurred a whole new conversation about due diligence done by investors into these companies while they're making investments into them.

[00:10:53] [SPEAKER_03]: So are investors being more diligent now?

[00:10:55] [SPEAKER_01]: Investors have certainly said that they've ramped up due diligence processes, asking for more information from startups, even doing background checks on executives and just putting the company under a microscope more than they would have, say, three or four years ago.

[00:11:13] [SPEAKER_03]: That was our reporter, Mark Vardabedian.

[00:11:16] [SPEAKER_03]: And that's it for Tech News Briefing.

[00:11:18] [SPEAKER_03]: Today's show was produced by Julie Chang.

[00:11:21] [SPEAKER_03]: I'm your host, Zoe Thomas.

[00:11:23] [SPEAKER_03]: We had additional support this week from Melanie Roy.

[00:11:25] [SPEAKER_03]: Jessica Fenton and Michael LaValle wrote our theme music.

[00:11:29] [SPEAKER_03]: Our supervising producer is Catherine Millsop.

[00:11:31] [SPEAKER_03]: Our development producer is Aisha Al-Muslim.

[00:11:34] [SPEAKER_03]: Scott Salloway and Chris Zinsley are the deputy editors.

[00:11:37] [SPEAKER_03]: And Fulana Patterson is the Wall Street Journal's head of news audio.

[00:11:41] [SPEAKER_03]: We'll be back this afternoon with TNB Tech Minute.

[00:11:44] [SPEAKER_03]: Thanks for listening.