On the Monday, July 27, 2026, episode of The Excerpt podcast: AI is already influencing decisions about health care, hiring, prices and more, yet the U.S. has no single federal rulebook governing the technology. Instead, oversight relies on a patchwork of state laws, voluntary standards, existing consumer protections and lawsuits after harm occurs. Sarah Myers West, co-executive director of the AI Now Institute, joins host Dana Taylor to discuss why regulators are struggling to keep pace, how the European Union’s approach differs and what an effective U.S. framework could look like.
Hit play on the player below to hear the podcast and follow along with the transcript beneath it. This transcript was automatically generated, and then edited for
clarity in its current form. There may be some differences between the audio and the text.
Podcasts: True crime, in-depth interviews and more USA TODAY podcasts right here
Dana Taylor:
AI is moving fast. Congress is not. In Europe, lawmakers have already passed the sweeping AI Act, but in the U.S., there's no single federal rule book. Instead, AI governance is taking shape through a patchwork of state laws, federal guidance, voluntary company promises, limited FTC enforcement, and lawsuits after something goes wrong. That may be the American way. Let innovation move first, then let courts and regulators sort out accountability later. But as AI tools move from answering questions to taking action, writing code, advising users, influencing decisions and interacting with people in vulnerable moments, is that enough?
Hello and welcome to USA TODAY's The Excerpt. I'm Dana Taylor. Today is Monday, July 27th, 2026. Joining me now to break it all down is Sarah Myers West, co-executive director of AI Now Institute, a nonprofit, nonpartisan organization producing diagnosis and policy research on AI. Sarah, thank you so much for being here.
Sarah Myers West:
Thank you so much for having me.
Dana Taylor:
Before we dive into regulation, I want you to just briefly touch on why so many of us, including tech companies themselves, feel it's so necessary. Can you lay out the risks we're all facing because of generative AI?
Sarah Myers West:
Sure. So AI systems have been in deployment going back decades, right? Different forms of AI have been used in a variety of ways from the very impactful areas of being used to shape whether or not a health insurance company is going to approve a claim that we make, whether we're going to get called in for a job interview, what's the cost of the goods that we're buying off of the shelves? There's many ways in which AI systems have already been in deployment and already shaping everyday people's lives. And exactly as you've said in your intro, by and large, regulators deal with the impacts of these systems on people's lives mostly after the fact. It's when a use of AI goes wrong, it discriminates against somebody or it makes an error that can cause some sort of harm, or it might be used for scams or fraud. That's the point at which we come in to remedy the systems. The current status quo is one in which people are getting impacted in ways that can affect their resources, their life chances, but without that deeper scrutiny that we need as the frontline of protection.
Dana Taylor:
Well, on an institutional basis, why can't the U.S. government regulate generative AI like any other technology?
Sarah Myers West:
Well, we do have a lot of laws already on the books that could be enforced for AI companies. I think lots of policymakers have said there really isn't an AI exception to the laws on the books. And I do think that that's a really good starting point. I think that, by and large though, we haven't seen that kind of front-footed scrutiny of this sector in the ways that are necessary given the impact that it can have on many people.
Dana Taylor:
As I mentioned, Europe has taken a very different path than the U.S. with the AI Act creating a broad, risk-based framework for AI. At a high level, how is Europe tackling regulation and is it working? And if it is, could the U.S. implement a similar strategy here?
Sarah Myers West:
So in the European context, regulators have passed this very large regulatory package called the AI Act. And what the AI Act does is it, by and large, breaks down different tiers of uses of AI by risk. It identifies that certain uses of AI are pretty low risk and thus don't really need very much scrutiny. Some are higher risk and domains that need a lot more attention. And there are some domains where the use of AI poses just an unacceptable risk and it can't be used at all. This approach I think has some benefits, some detractions. I think one of the benefits is that it allows for that attunement to the actual context of use. One of the detractions though is that a lot of AI is not developed for a particular use case, but is general purpose technology and thus needs more scrutiny before it's out in broad use. In the U.S. context, we largely have not taken that kind of broad strokes approach to the AI sector, but it is one of the paths forward that some policymakers have considered.
Dana Taylor:
So instead of the regulation that we're seeing with AI in Europe, we have three states with similar laws focused on frontier AI safety and transparency, California, New York, and the most recently Illinois. And then there are very different laws in Colorado and Texas. How do these states try to mitigate risks and are they enough?
Sarah Myers West:
So there have been a variety of different states that have passed laws that attempt to wrap their arms around this complicated issue area. One set of approaches focuses on trying to mitigate the farthest out catastrophic risk. This is the type of law that's been passed in California, in New York and elsewhere. I think one of the limitations of that approach is that, by and large, it exempts most of the uses of AI that are already in deployment from scrutiny. It only applies to AI systems that are above a certain size and scale. That means that most of the harms that people may be feeling in the here and now aren't really getting very much attention. It's a mitigation against future-facing, more catastrophic risks. But it can be equally existential if AI is used, like I said earlier, to deny your health insurance claim. And that is an area that needs much more attention.
Dana Taylor:
NIST, the National Institute of Standards and Technology, created the AI Risk Management Framework, basically voluntary guidance for managing AI risks. In a country without one binding AI law, is NIST still the most important place where AI safety standards are being shaped? And then what about the Federal Trade Commission and the Consumer Financial Protection Bureau? What roles should they be playing?
Sarah Myers West:
So as you mentioned, the National Institute for Standards and Technology has historically played an important role in the tech sector in providing evaluation frameworks for assessing AI systems. And one such framework is the Risk Management Framework. These are really designed in ways that are meant to just inform businesses but are not binding. They don't have the kind of friction that's going to force companies to behave in certain ways or change the incentive structures. And I think what we're seeing now in the dynamics around AI is a lot of companies being allowed to grade their own homework. And that is not going to be meaningfully protective of the frontline impacts that communities are feeling across the country.
Dana Taylor:
President Donald Trump entered office promising deregulation along with AI dominance globally. His administration has also exerted pressure against state AI laws. Meanwhile, his key AI policy advisor, Sriram Krishnan, resigned at the end of June. How should we understand the direction of Trump's AI policy?
Sarah Myers West:
The Trump administration's stance toward AI has been, if anything, to take an already stepped back approach to this sector to the farthest extreme. It has both attempted to create deepened regulatory carve-outs that inhibit the ability of states to use their jurisdiction to protect their own constituents. And then equally has taken this approach where the administration has doubled down on providing support to particular businesses, both through trying to broker deals internationally on behalf of particular AI companies under the Export AI executive order, allocating federal lands for data center construction, providing billion-dollar loans to some AI companies for data center build out. So I would say that this is an administration that has really put its thumb on the scale in favor of certain interests within the AI industry.
Dana Taylor:
Does Congress need to step in to empower federal agencies like the FTC and NIST to regulate generative AI? Is that the solution? And if so, is there a political will to do it?
Sarah Myers West:
There is certainly an important role for Congress to play here. One is ensuring that there's adequate technology staffing and resourcing to be able to enforce the laws that we already have on the books before they even get to strengthening legal frameworks that enable us to have meaningful oversight and accountability. I think one of the most significant gaps is that, given the major economic and political capital that AI firms have amassed, these are some of the largest and most well-resourced companies that we have seen really ever. It makes it very hard to apply the law in a way that's going to create real friction for them. For a lot of these companies, an FTC fine can be a line item that they provide for in a budget. And so I think it's really crucial that congressional leaders not only act, but act in ways that are going to be meaningfully changing the behavior of firms that have demonstrated themselves not to be thinking about the needs and interests of the broader public.
Dana Taylor:
Sarah, when AI causes harm, how do U.S. courts and regulators decide who's responsible? The company that built the model, the company that deployed it or the user?
Sarah Myers West:
It remains an area that is somewhat vague. And it can be tricky with the generation of AI systems that are in deployment right now, because a lot of these systems are designed in such a way that it's hard to attribute responsibility. It's hard to say definitively, "This input led to this output. We know that, because it was trained on this data, it caused this discriminatory outcome." Some of this is by design too. These companies used to publish out research that gave the broader public information about what kinds of systems they were building, what data they were trained on, things that are really important for the broader public interest. They become increasingly opaque to the public and to regulators as time has gone on, making it even harder for courts to be able to enforce the law.
Dana Taylor:
And then finally, what might a workable U.S. AI governance framework look like, one that protects people without discouraging innovation?
Sarah Myers West:
I think we need to dismiss this fiction that regulation harms innovation. I think under the current paradigm, there's a lot of regulatory oversight that would really help spur on innovation. It would help combat the monopolies that have developed within the sector. It would protect the public. It would also protect businesses from untested systems that might have security flaws that might not work as intended. There's a lot of ways that greater regulatory scrutiny would really be beneficial and lead us to developing AI that's at the gold standard by weeding out some of the bad apples that have developed along the way.
Dana Taylor:
It was good to hear your insights. Thank you so much for joining me, Sarah.
Sarah Myers West:
Absolutely. Thank you for having me.
Dana Taylor:
Thanks for listening. I'm Dana Taylor. You're now caught up in under 20 minutes. Come back tomorrow for another take on a story that matters.
This article originally appeared on USA TODAY: Who regulates AI when Washington won’t? | The Excerpt
