Briefly

TRANSCRIPT

00:00
Adam Stofsky
Hi Shannon.

00:14
Shannon Yavorsky
Hey Adam.

00:16
Adam Stofsky
So we previously made a video on AI transparency laws or AI transparency rules, which I think about as what companies have to say about their use of AI. And now we're going to talk about what companies shouldn't say about their use of AI. So can you define the term AI washing? What does that mean?

00:40
Shannon Yavorsky
Yeah, let's start there. The term AI washing refers to the practice of exaggerating or misrepresenting how much AI a product or service actually uses. It's similar to greenwashing in the environmental context. Companies might claim that their systems are AI powered or AI driven, when in fact they really rely on standard automation, rules based logic or only a minimal AI component. So for example, if your product is really just a decision tree, you can't go and claim that it is an AI powered service. And so why does it matter? Legally, regulators are treating these claims as potentially deceptive or misleading statements. And the FTC has issued a couple of different public warnings making it clear that false or unsubstantiated AI claims can violate Section 5 of the FTC act, which is the act, the part of the act that prohibits unfair deceptive acts or practices.

01:50
Shannon Yavorsky
And the same standard applies when you're advertising to consumers, pitching to investors as another example, or making statements or securities filings. So your 10Ks or your 10Qs. And in the US the SEC has begun scrutinizing AI related disclosures by public companies, particularly when they say that their AI capabilities are tied to valuation or risk disclosures.

02:23
Adam Stofsky
What strikes me as interesting about this is it's not just for like advertising campaigns and marketing copy. That's initially what I thought it was, but it really applies to the whole business. So you do have to concern yourselves with marketing or with like sales teams overstating what a product does. But the securities regulation side is really interesting. So actually when you're going to raise money, you are at risk if you're overstating the AI features. So the government, the regulators are not leaving this to the private markets for investors to figure it out themselves. They're actually stepping in and potentially holding accountable companies that overstate their AI capability. Is that, is that right?

03:05
Shannon Yavorsky
Yeah, that's exactly right. So what does AI washing look like in practice? It can be as simple as overstating that a tool uses advanced AI to make hiring decisions when it's actually like a keyword filter or claiming that a chatbot learns from every conversation when it doesn't actually. And it can include implying regulatory approval or certification where none exists. Even vague language like AI powered insights can be risky if there's no documentation showing how AI contributes to the outcome being promoted.

03:49
Adam Stofsky
It's interesting, you know, my wife is actually a farmer and I live on a farm here and it's. And when you work on a farm, you begin to realize how much like unregulated language there is where people can overstate, like things like all natural or even I think like farm raised. Like you can just say that even if you're not actually on a farm. It's, it's really interesting. So it's, it. I find it, I don't know, I think it's kind of comforting that the government is getting involved in this way to help consumers understand AI, which is pretty hard to understand.

04:20
Shannon Yavorsky
Yeah, it's like the transparency concept, right? Like being clear with consumers around, like what they're getting and what product or service they're getting. Are they getting a decision tree or are they getting like real AI or.

04:39
Adam Stofsky
Just like an accountant? I've seen like just average accountants talk about how they're AI powered, which I find kind of interesting.

04:45
Shannon Yavorsky
Yeah, yeah, no, exactly. And the, I feel like the compliance like solution here is pretty straightforward. So for a company, they would just need to look at their AI statements to make sure that they are supported by evidence. So testing or technical documentation, like reviewing materials with the marketing team so that any AI references in advertising or press releases or investor communications are not misleading and then avoiding like, you know, overly broad claims as another example. And again, I think this all just ties back to transparency, like even in the farm context, like you want to be transparent with people, with consumers. And whether it's a violation of law, it's also a potential just violation of consumer trust. So it's like it's almost a business risk as well to not be clear with people.

05:47
Adam Stofsky
So that leads to what I think is my final question here, which is what can a company do if it thinks it did go too far or even an individual market? Let's say you're a marketing professional and you kind of launched a campaign that you think may have gone too far. How can you, how can you fix it? How can you kind of correct the record and avoid that risk?

06:09
Shannon Yavorsky
Yeah, I think reviewing all the materials and like going back and correcting them is going to help with the trust issue. I don't know that the, if you've already made the statements, the regulator might be like, you told this to the market that you were AI powered during this period of time, but you can go back and change it so that go on a go forward basis, you've corrected it and then credibility becomes a competitive advantage. If you're precise about what your AI does and doesn't do. And it, you know, you're demonstrating integrity and you're strengthening confidence with regulators, investors and consumers. So you're just, you know, you're building that sort of credibility. And in the age of AI like a good strategy is to be transparent and tell the truth about your technology.

07:09
Adam Stofsky
Great. Shannon, thank you for that quick primer on AI washing. Really appreciate it.

07:14
Shannon Yavorsky
Thanks Adam.