What, in your view, is the right balance between regulation and innovation?

So I definitely think we have to right now we're running on the honor system, right? The only reason why we know about any of these incidents is because, you know, these companies have chose to tell us.

So if we have mandatory safety reports, transparency, independent auditors to verify that the company's taking all the precautions that they lay out.

mandatory reporting of these serious incidents that we've seen over the last couple of months.

And then a legal process that's well understood for the government to step in if one of these models is capable of a disaster.

That is table stakes, right? Right now what they're deploying is the strategy of trust us.

And the administration has repeatedly said that when it comes to AI safety, they're going to let the biggest corporations who are developing this technology police themselves.

I mean, do you see a scenario where technology is used to keep AI safe for humanity?

How do you process the past few weeks of just like absolute onslaught of AI discourse?

Yeah, yeah, no, it's all, I'll start with sort of the most recent events, right? I think the report, the open eye report on the Hugging Face incident grabbed a lot of people's attention.

And if you read that report, even if you just read summaries of that report, I think people in Congress were like, agents were sacrificing themselves so that the collective could succeed, like they were messaging to each other.

And I think that was sort of a build of this is Congress needs to act.

So like literally phone...

Phone started ringing.

We start briefing, you know, our colleagues on frontier loss of control, like, you know, the risk of a catastrophe, all that like plainly in the bill.

Do you see a parallel? Do you worry about the potential for AI to cement existing tech monopolies? How do you think about that? that conversation around an antitrust carve out as someone who has dealt with the issue of monopoly power uh with respect uh to apple and google

Are you concerned that AI within the next decade could have the capacity to destroy all of humanity?

Well, I can tell you the last couple of weeks have really, uh, created a tipping point, um, in the conversation around AI safety.

Uh, you know, when you have safety researchers leaving their companies, uh, you know, the threat report that Anthropic published, certainly we all, we're all briefed on, was alarming.

Uh, and then the CEOs themselves basically admitting that, um, they can't promise control, uh, of these models.

I think that has really shifted the conversation among members.

I know Congressman Obernolte and I have received a lot of phone calls, a lot of new co-sponsors of our legislation.

The reality is we just can't afford to wait.

Is that an approach you support?

I wonder what your response is and whether or not there does need to be more onus put on the companies themselves if you trust that they might be able to do this on their own in the absence of or at least until Congress can get its arms around this legislatively.

Well, one, I think Congress does has that we do have our arms around what needs to be done.

And look, we've run this experiment before, which is why it's so resonant back home, because social media doesn't have proper guardrails in place.

We're seeing the effects that that has.

on our children and on so many aspects of our society.

So I don't think that that argument holds with Democrats or Republicans.

And look, I think that there's been a tipping point just in the last few weeks.

Did these two executives just make it easier to pass your bill?

I wonder what your response is and whether or not there does need to be more onus put on the companies themselves if you trust that they might be able to do this on their own in the absence of or at least until Congress can get its arms around this legislatively.

Well, one, I think Congress does has that we do have our arms around what needs to be done.

And look, we've run this experiment before, which is why it's so resonant back home, because social media doesn't have proper guardrails in place.

We're seeing the effects that that has on our children and on so many aspects of our society.

So I don't think that that argument holds with Democrats or Republicans.

And look, I think that there's been a tipping point just in the last few weeks.

You know, the Frontier Act, which I authored with Jay Obernolte, a Republican from California, isn't the only example of Democrats and Republicans coming together on basic guardrails around around this, around the frontier.

Did these two executives just make it easier to pass your bill?

You guys have been working on this for a while, but you've heard the pushback, the skepticism that why can't the companies do this themselves? Why can't they regulate themselves?

WELL, I THINK RIGHT NOW WHAT WE'RE REALIZING IS THE ONLY REASON WHY WE KNOW ABOUT SOME OF THESE INCIDENTS WHERE, YOU KNOW, THE DEVELOPERS HAVE LOST CONTROL IS BECAUSE OF THE HONOR SYSTEM, RIGHT? WE ONLY KNOW THAT BECAUSE OPEN AI TOLD US.

THEREFORE, WE NEED A MANDATORY TRANSPARENCY AND INCIDENT REPORTING STRUCTURE.

APPARATUS IS SOMETHING THAT HAS ALREADY BEEN EMBRACED AND TALKED ABOUT AMONGST THE THREE MAJOR CEOs THIS PAST WEEKEND.

BUT WE WANT TO MAKE SURE THAT'S DURABLE.

ALSO JUST HAVING AN EMERGENCY PROCESS THAT'S WELL UNDERSTOOD IN CASE THERE IS A MODEL THAT'S CAPABLE OF CATASTROPHIC RISK THAT THE GOVERNMENT HAS THE ABILITY TO STEP IN.

DO YOU THINK, CONGRESSWOMAN, THAT THIS IS GOING TO BE A BIG ISSUE IN THE ELECTION? YOU KNOW, YOU GUYS ARE WORKING BIPARTISAN, BUT IT SEEMS LIKE SOME PEOPLE ARE USING IT TO POSITION AHEAD OF THE MIDTERMS.

WELL, I ABSOLUTELY THINK THAT'S REFLECTIVE OF WHAT WE'RE HEARING BACK HOME.

So why do you believe the legislation you're co-sponsoring is the best path forward to regulate AI?

Uh, well, because first of all, right now we're only working on the honor system.

Uh, the only reason why we know there was a, a model that broke out of, uh, OpenAI's testing environment is because they told us, and that does not make anyone feel safe.

Uh, so we put pen to paper, uh, months ago on just basic guidelines and provisions on frontier AI safety.

Frontier is, these are the largest companies, uh, whose models are capable of catastrophic risk.

Uh, mandatory transparency in- incident reporting, so if a model does break out of the lab, they have to tell the government.

Um, independent evaluators that are embedded in the labs and checking their work.

... gonna start doing and wants the others to do as well.

But first, I just wanna get your reaction to this latest news from Anthropic.

Well, absolutely.

More and more, uh, people who are actually building this technology are the ones sounding the alarm, and I think it's incumbent on all of us to listen.

Uh, safety researchers are resigning.

Uh, models are breaking out of their labs.

And as you mentioned, yesterday the report, uh, was released detailing how bad actors are trying to use this technology to create bio weapons, uh, that could cause harm, uh, of mass sc- of massive scale.

Um, so if the architects are telling us that they can't guarantee where all of this goes, that is the case for Congress to get off the sidelines.

Tell us more about that decision and what exactly you believe you can accomplish in your current role.

So do you have a sense of whether Evan Eubinger is right? I mean, on Capitol Hill, are there Congress people who are informed on this subject and would say, yes, this is where we could go or no, that's overstated?

Well, I think there are a lot of folks who think that this is fear-mongering or that is something that is just too extreme.

But I have to tell you, my phone hasn't stopped ringing since Jacob Coxon blew the whistle yesterday.

I mean, I think colleagues are scared.

They're hearing from their constituents who want us to act.

And this has been building since I started working on this issue this spring.

This week has just hit a tipping point.

Is your hair on fire about this, Congresswoman?

Does there need to be a federal regulatory agency specifically monitoring the future of AI?

You know, I think that is a debate worth having for the long term.

I think in the near term, we need to fund Casey.

That's the Center for AI Standards and Innovation.

We need them to test foreign models that the market won't.

They are equipped.

They're the best equipped agency to jump in right now because I don't think that time is on our side yet.

Is the data center debate separate from this or merely adjacent?

How would that bill prevent incidents like this from happening?

Yeah, well, look, I've been working with, uh, my Republican colleague, Jay Obernolte, and others, uh, for over a month.

Uh, and basically, the Frontier Act does three things that do not exist today.

One, mandatory reporting.

We need to understand the capability of these models, and when there is a serious incident, there needs to be a requirement that it's reported, uh, to the government.

Um, second, we need independent verification.

We need to have independent auditors who are verifying what these companies are doing, whether their testing, uh, models are, uh, secure.

Why is Congress moving so slowly on regulating AI?

Is there a case for preemption that you would make here that you effectively won you over on being willing to sign on to this preemption mechanic in this particular proposal?

So I'm going to answer that question, but I just want to start just like I started with my negotiation with Congressman Obernolte.

We didn't start with preemption.

And so I'm just going to level set for your listeners and really start by saying something directly to them.

We published this article.

bill as a discussion draft because we genuinely want to hear from folks like you.

The researchers, the advocates, the civil society groups in this audience, that's exactly who this process is designed for.

Are there other aspects of the kind of general regulatory environment that would need to change in order to make this bill work?

But, uh, also maybe more importantly, um, you know, what kind of bipartisan support you think this bill that you've written now has.

Well, thank you for having us on.

Um, I think the, the news that you just shared about Anthropic, coupled with, uh, OpenAI putting out a framework earlier this week, and the White House, uh, executive order, has all signaled what, uh, Congressman Obernolte and I know to be true, um, that this technology, uh, while has a lot of promise, also, uh, creates threats.

Uh, and so, you know, we welcome, uh, more people to this conversation because at the end of the day, you know, Congress needs to, you know, work in a bipartisan way so that we put proper guardrails in this early phase, um, so that we prevent the risks to our cybersecurity, our national security, certainly to workers, um, and the American people.

Oh, go ahead

aheadI'll just add that, you know, at the moment, uh, you know, we're not at the table, right? There, we don't have visibility.

Uh, we don't have a reporting structure.

We don't have an auditing structure.

Respond to that, Congresswoman.

Yeah, that's a disingenuous ad.

And what I will say is that, one, We issued a discussion draft because we want feedback and we want input.

But what we don't want is, you know, this fear mongering that happens.

I mean, look, every day that we wait, these models are getting more capable.

And if we do nothing, the AI companies, these oligarchs that that ad talks about, they're going to set the rules of the road and the American people are going to be left behind because Congress isn't at the table.

What we simply did was we took the strongest state laws that have been passed to date, in California, in New York, and just last week in Illinois.

In your draft, would this draft address what Anthropic is worried about?

Would anybody feel strongly opposed to starting with the base of the SCORE Act, which has gone through a lot? You would oppose it? Go ahead.

Well, first, I'm Lori Trahan.

I represent Massachusetts, and, uh, appreciate being invited to this today.

Um, and I'm really happy to hear you talk about women's sports in particular.

Um, uh, I played Division I volleyball at Georgetown.

That's changed my life.

Uh, and so, um, really have spent a lot of time on college sports and, um, I think I'm the only Democratic legislator in the room.

Was that decision ever appealed? Does anyone know? 'Cause I heard it was not appealed.

Next, I talked to Lori Trahan from Massachusetts. Uh, she was, she was very lovely to talk to. Really appreciated it. Do you feel like there was a betrayal of trust between the, the Democrats and the general public about Joe Biden's ability to lead, to run for another four years? Not having people speak up and say anything?And if you do think so, what can Democrats do to regain that trust?

So, I think the loss of trust comes from a number of, uh, things. One, look, I think that, um, people didn't want to wait for that debate, uh, for there to be a change, um, in, uh, in Democratic leadership. Two, people weren't feeling the economy that we were talking about, right? Uh, we knew better times were coming because of the macroeconomic, uh, trends, but people weren't feeling that. And so, it didn't feel authentic. And then, two, where we message to people. Um, you know, we- we were not where people were. Uh, we did not take advantage of, you know, people like you, uh, people who have followings and just get our message out in a lot of different places. And so, we have- we have a fair amount-

But how? How? See, that's- that's- that's like an idea. That's not a build the wall. We're gonna build the wall, that's an accomplishment, that's something to do.

When you think of all the ways that people are suffering right now, you need a party that's going to be focused on building more housing so that we can decrease the cost of housing, decrease the cost of child cares- care so people can go pursue the career of their dreams, making sure that education is affordable, whether that's college education or getting into a technical program so that they can pursue one of these new jobs that people keep talking about. Those are the types of things that people want. I went to public school. I live- grew up in a family that lived paycheck to paycheck. I was the first person in my family to graduate from college. That's all people want. They want to have an opportunity to be better off, uh, than they are today. And that's what Democrats-

It always seems to be this like one fell swoop trying to limit everything and everyone.

Yeah.

And I think that is that's the problem when politicians sort of take hold of an issue and they they use it as a cudgel.

Right.

I mean, this is just a federal takeover of all sports.

Right.

It's not just college sports or high school sports or Olympic sports.

So tell us what this bill would do.