LIVE: House Democrats Hold Press Conference on AI

Live News Network
1.4K followers
23 Video Views·Sep 16, 2026
House Democrats hold a press conference on artificial intelligence at 9 a.m. ET on Sept. 16 in Washington.
AI Generated·May contain errors
Video Summary
AI GeneratedA bipartisan group of members of Congress is urging immediate federal legislation to regulate artificial intelligence. They argue that corporate self-regulation is insufficient and that the government must establish safety guardrails and independent verification to prevent catastrophic risks, criticizing House leadership for inaction and prioritizing recesses over urgent safety bills.
We're thrilled, especially if you stayed with us until 11 o'clock last night.
Okay, I'll come back this morning.
We're here today as part of a bipartisan group of members of Congress,
and in Washington, and certainly blowing up from our constituents,
that we need to pace the AI frontier in ways to pursue the societal benefits of artificial intelligence
while making public safety and security paramount.
There's so much more important than a couple more weeks of congressional recess,
and many of us in both parties want our leaders to tackle these in a serious way
before Congress adjourns for two months.
Congress has a history of only responding to disasters only after they happen,
only when there's way more than clear warning signs.
I think all of us can think of examples where Congress has had emergency votes to respond to a crisis.
We cannot wait for the AI to act,
and there have been too many people in the AI world who thought that it would take a disaster to actually get us to act.
But we're in a different place.
We can't wait for that catastrophic failure of leadership.
I'm hearing from my constituents that in what they're hearing on AI risk,
from Mythos to the Hugging Face incident to Cox's, to Bill Gates, Barack Obama, and all the leaders,
there's this tipping point where, especially seeing all the researchers and the safety staff
at places like Anthropic and OpenAI talking about the dire consequences of this technology.
They want to know that they're safe.
They want to know that their kids and their grandkids are safe.
And while we appreciate the private sector's willingness to voluntarily slow down model development,
that's not all of the private sector.
And so if we look, we don't be naive enough to think, oh, self-regulation is just going to work.
We don't allow that in virtually any other sector of our government.
We don't allow it in the SEC.
That's why we have NHTSA.
That's why we have FTC and the FCC.
This cannot come just from the executive branch either, though,
because as we have seen, it's failed to be transparent about its proposed AI testing framework.
Donald Trump says he has one, but we have no idea what it does or whether it will be there.
And Congress must act and pass legislation.
We have bills, many bills, that have broad bipartisan consensus.
And that will be a starting place.
But we also must enact broader regulation to address systemic safety risks.
The details matter. We get this right.
We read this morning about the big Democratic letter on AI.
Most of us have signed it. We approve.
But we also know that this has to be a bipartisan effort,
which is why we've also recruited a number of very important AI Republican leaders in the House
to join us in this effort.
With that, let me turn it over to a leader on the Energy and Commerce Committee,
the author of one of the most famous current pending pieces of legislation,
and a great thinker, and also the mother of the first fusion energy plant
in the history of human kind.
Great introduction. Thank you, Congressman Beyer.
Good morning, everyone.
It's great to be here with you, Congressman, Vice Chair Liu, Congresswoman Jacobs,
three of the most serious voices in Congress on AI.
Look, tomorrow the House is scheduled to leave town until after the midterms.
That's 53 days. That's how long we'll be gone,
53 days before Congress can take up legislation to put guardrails on AI.
Nothing in the meantime to address the serious risks detailed by Jacob Coxon
and other safety researchers like him,
nothing to mandate the placement of independent verifiers inside frontier labs
to evaluate the dangers their models pose,
and nothing to make sure that when a model poses an imminent risk,
the federal government has a clear, nonpartisan, court-backed process
for halting its release.
Instead, a chorus of leaders here in Washington have decided
that they are comfortable gambling the safety of the American people
by arguing that we should pass no guardrails at all.
They've actually said that to do anything,
even basic transparency in incident reporting standards,
would cede America's leadership on AI to China.
Yet, the same folks who oppose AI regulations by citing a race with China
are silent while the president greenlights the sale of American-made advanced chips to China.
And they say nothing while the White House maintains a de facto licensing regime
negotiated under complete secrecy
and arbitrarily applies export controls to leading models on a Friday night.
The argument just doesn't make any sense.
And if this administration is serious about leading the world on AI,
it would work with Congress to set the rules of the road for this technology,
just as we have with the most dangerous technologies throughout history,
nuclear weapons, chemical and biological weapons, aviation safety, just to name a few.
But to do that, we need lawmakers here in Washington working to hammer out a deal.
Each of us here this morning has proven that we are willing to work
in good faith with Republicans, Republicans who understand the same thing we do,
that to unlock the tremendous promise of this technology,
we have to do everything possible to mitigate its accompanying risk.
We have legislative proposals ready.
We just need the willpower to move them forward.
So our ask is simple.
Keep us in session. Bring us back to a special session.
Or at a bare minimum, keep a group of us here in Washington,
people like Don, Ted, Sarah, and me, as well as our GOP counterparts,
who have a track record of working productively across the aisle on this issue.
Let us get this done so that we can take it up as soon as possible.
The families we represent don't get a recess from this technology.
Neither should we.
Thank you, and it is now my pleasure to turn it over to the vice chair of our caucus,
Congressman Ted Lieu.
Congressman Don Beyer and Laurie Trahan and Sarah Jacobs,
I note that last month U.S. gas prices on a monthly average set a record high.
This month, diesel hit a record high,
and now we have three leading AI company CEOs saying we need to slow down AI development
because they understand the products they're making are not safe.
And what is Speaker Johnson's response?
He literally changed the House calendar to give us two more weeks of vacation time.
That is completely out of touch with the American people.
It is a wrong decision.
And last year, he made a specific decision to terminate any task force,
any kind of group working on AI,
even though last term there was a bipartisan House task force on AI.
Speaker Johnson chose to discontinue that, which is quite legitimate.
It is a good policy decision to create a democratic AI commission and task force.
So Leader Johnson has had a long track record of sticking his head in the sand
and ignoring the AI risks that we're seeing now.
And I also just want to make a comment about some of the folks in the industry.
Some of the leading AI CEOs understand what they're seeing.
They understand the dangers.
Others are talking about how market forces can fix this.
We don't need new laws and regulations.
And my response is, who do you think you are?
Market forces did not cause airbags to be in cars.
Market forces failed to cause seatbelts to be in cars.
Market forces failed to allow drugs to be safe.
Market forces failed to prevent acid rain.
That's why the federal government regulates vehicles and regulates drugs and regulates power plants.
So who do these AI folks think they are, these elite folks?
They are somehow special.
Well, they are not.
They are creating products that are not safe.
And we here at Democratic Caucus are saying, keep us here in Congress so we can regulate
and make sure that these products, in fact, are safe.
Now it's my great honor to welcome from Southern California, San Diego area,
and it's an honor to be here with you.
Well, thank you, Don, Ted, and Laurie, and thank you all for being here.
I've spent over a decade working on a UNICEF representative to the AI for Good Summit in Geneva in 2017,
before it was cool.
And what that decade has taught me is two things,
make the world a better place.
And it can hurt people in small, everyday ways and in the catastrophic ones.
And Congress hasn't done enough to make sure that it's doing the right things right now.
Ask any young person, and they'll tell you.
For the first time, CUBE research found that most adults under 30 are more concerned than excited about it.
And yet, President Trump, keep repeating the same bad faith arguments that if we regulate AI, we'll lose to China.
It's a complete lie.
And it's a fundamental misunderstanding of how and why people use technology.
The race with China isn't only about who can build the most advanced model first.
Getting to the best system first doesn't matter if people don't use it or if AI has killed all of us in the meantime.
Winning the competition is about which models people trust more to use and adopt.
Putting guardrails on AI, stopping the everyday harms from happening when an algorithm gets to decide who gets a mortgage or a job interview,
and getting ahead of the bigger risks before they arrive, that isn't losing.
We win it.
The FAA did make us lose the competition on air travel.
The Clean Air Act helped car companies learn to innovate with catalytic converters.
It didn't get in the way of their innovation.
261 scheduled voting days.
That's almost a quarter of the days that Congress could have been in session to pass any one of the numerous bipartisan AI bills in the chamber.
It's 62 days that could have moved the 89 recommendations that the bipartisan AI task force identified almost two years ago
to guide federal AI governance and innovation.
But instead, Speaker Johnson comes and preaches that big tech can police itself.
It can't.
It's our job, and we're not doing it.
Enough is enough.
The American people have way too much impact on AI, and we can't let all these warning signs pass without action.
Speaker Johnson needs to cancel recess so we can get to the hard work of Congress
of getting to real consensus, hammering out the details, and passing real legislation.
Thank you, Sarah, very much.
Obviously, we have great folks here willing and ready to answer your questions.
Yes, please.
I spoke to some of your colleagues on the other side of the aisle yesterday.
A lot of them said there's no role for Congress here or limited role for government here.
What's your take on that?
And you're trying to kind of move forward and say you're willing to stay for days when Congress is out of session.
Yeah, I don't think it's an option to abdicate our responsibility and to go home,
nor do I think it's an option to abdicate our responsibility to tech executives to regulate themselves.
We've seen that movie before where those of us with children are actually living it with social media.
There's no life for us to stay and continue working on the legislation that we've written
and this AI frontier lab safety.
I'd like to pile on, too.
Congressman Lew pointed out seatbelts and airbags and acid rain.
Let's just say as somebody who made my life as a car dealer, from 1992 to 2008,
the CAV-A standards stated 22 miles per gallon.
And so every change in the automobile went to make it bigger, heavier, faster.
And when Barack Obama came in, they moved it to 34 miles a gallon.
It took a decade.
Markets do not work in many, many different ways.
Another question.
Yes, please.
How do you plan to reach consensus on this when you can't reach consensus among your own caucus?
You have a bipartisan bill by Representative Trahan and Obernolte.
That's a bipartisan bill.
It's ready to go.
You know, Vice Chair Lew, why not just act on that if it's such an urgent issue?
Yeah, so let me commend Congresswoman Lori Trahan and Congressman Jay Obernolte for introducing legislation.
And they introduced a discussion draft, and then they did what's normal in the legislative process.
They took input from lots of different stakeholders, and then they revised it.
And I absolutely believe that the Energy and Commerce Committee needs to do a markup of that bill.
It wouldn't go on vacation in a couple days because that was Speaker Johnson's choice.
By the way, I thought what Sir Jacob said was quite interesting about nearly a quarter of the session days being canceled by Speaker Johnson.
It's almost like he's filibustering his own speakership.
It's quite amazing to watch.
So do you support the bill, or what do you need to support the bill?
Well, I support the legislative process.
There's lots of different bills, and I think they should be heard in committee and have markups.
And I also think committees just need to do hearings.
But you guys started the press conference saying this is like an imminent threat.
So if it is such an imminent threat, why not just get on board?
You know, that's a regulatory framework that both Republicans and Democrats can work with.
Get that across the finish line.
Yeah, so something I'm thinking about is I should just go out and co-author all 37 bills that I've introduced on AI.
I might just do that.
Yes, please.
Good morning.
Thank you for having this.
One of the larger partisan divides on AI legislation is going to be preemption.
And I'm wondering about the posture that the four of you, maybe you can speak broadly for the House Democrats, right,
may have going into negotiations with your Republican colleagues.
I'll say I don't know that the four of us all share a sort of common view on it.
And I think that we haven't even been able to do the real legislative process of finding consensus
because so many of these session days have been canceled.
None of the committees have held real hearings on this.
Like, we need to do that work, and I believe it is possible for us to get to a real consensus that we can pass through the House.
I can't support California.
I can't support anything that will mean that my constituents will have less to do today.
And we know that the state of California, because Congress has failed to act, has enacted a number of numerous proposals.
Now, that doesn't mean that there's not a world where if we get really good legislation at the federal level,
that will go into effect immediately, that there's not a conversation to be had about preemption.
But the way I think about it, at least, is that I can't let my constituents do what they're doing right now.
And that is kind of the framework I think about as we're going into these conversations.
I'll just say something about preemption, because I think that question is probably a direct result of the process that Congressman Obernolte and I went through.
Clearly, we put out a discussion draft that dealt with a lot of titles, went across a lot of different sectors and committee areas,
and we chose to take the feedback that we heard from all stakeholders, but, you know, researchers, universities, student groups,
consumer advocates, phasing our Frontier Act, which is really focused on catastrophes and should occupy a federal lane.
As part of that process, we tightened up the preemption so that it was only on the very things that we were putting in.
Incident reporting, that's mandatory, which, by the way, we don't have today.
The only reason why we know about the Hugging Face incident is because OpenAI chose to tell us.
Independent verification and an emergency process that the government can step in if there is an imminent or if there is a risk of a catastrophe.
In a perfect world, if I had my way writing that bill, would it look different? Absolutely.
I choose to legislate in the reality that we're in right now, which is we're going to need this to be bipartisan.
If we're going to pass it tomorrow, we will likely need it to be bipartisan.
Even if we win back the majority in November and get something done in the lame duck, we will also need it to be bipartisan.
I think that this moment warrants us coming together and being really smart about what can pass because the risks.
I think we've had the warning shots, right, that a lab is going to cause real damage.
So that's sort of how I think about it.
Certainly, we will see how things change over the few months, and we'll continue to negotiate with our colleagues so we can get something across the finish line.
Thank you.
Yes, please.
You've all spoken about the lack of urgency here in the House.
I was wondering about the Senate and if you are seeing urgency.
You think of there where some of the leadership is involved, centers like Boone, Cruz, and Klobuchar.
To what extent you all are directly engaged in conversations with them to try to coalesce your efforts to actually get something done?
I'm very pleased that the United States Senate is introducing legislation on AI, and a number of legislators on the Senate side have introduced legislation much earlier than recently.
I also do want to commend Republican Senator John Kennedy for introducing a version of the AI Kill Switch Act because I believe that the first principle of AI safety is humans have got to be in control and not the other way around.
And at minimum, that means that we need to be able to shut down AI agents and models that go rogue or have advanced capabilities.
You have Senators Josh Hawley and William Methow also introducing bipartisan legislation.
They're re-upping legislation on AI regulatory issues.
So there's a lot of people working on these issues.
And then the Senate did have sort of an AI task force that is in session at the end of this month, so should the House be.
Let me add, obviously, Senator Sanders is working with Greg Kassar on some of the paws that are facing the frontier.
And Mark Warner has been leading on AI for a long time, specifically looking at the job impacts.
And we are all talking to all of them all the time.
What do you guys make of the argument that this doomsday scenario that these companies laid out was surely for their own self-interest?
To protect themselves?
In a world in which we live.
But I heartily reject that.
I think if you read carefully about Demesis Obvious going back to the 1970s and Dario Amadei and Altman and others,
they all, Jeff Hinton, they all started from a premise of safety.
They're all aware of risk from the very beginning.
And, in fact, DeepMind, OpenAI, and Anthropic were all literally founded on the concern about risk and about safety.
So it's been part of their culture from the very beginning.
So maybe it will have a financial impact.
It could also have a negative financial impact on it if they actually face the frontier or slow down.
And I also think when you look at the many, many people in the organizations,
the people that are part of those things are there for the right reasons.
They're not there to make a lot of money.
So let's just briefly talk about the hack-a-hugging phase.
You had OpenAI basically create 1,200 AI agents that they put in sandboxes where they're not supposed to escape.
Imagine it's a locked room, and they gave these agents a cybersecurity exam and say, complete it.
These agents figure out a way to break out of that locked room.
And then they manage to find a message board that they use at OpenAI,
and they start communicating with each other.
And then they formed what is known as a collective.
They elected, essentially, an AI agent leader.
They had AI agents then go out into the Internet to steal the answer key to the cybersecurity exam.
About 700 or so of these agents executed about 16,000, 17,000 transactions and attacks on HugMeFace.
They penetrated HugMeFace, another entity.
And then they had Kamikaze agents that would go and essentially commit suicide
for getting information back to the collective.
And you think OpenAI made all this ---- up as a regulatory capture scheme?
Then you are just blinded by profit and partisanship.
These people understand that they're creating products that they've lost control of and are not safe.
I think some of the concerns that people have around legitimate concerns,
and I think we need to take them into account as we're thinking about what Congress does, right?
We don't want to create a regulatory structure that actually makes it harder for open rate models, for instance,
because I think, you know, as we look at the competition with China,
open weight models and investing in American open weight models
is going to be a huge part of how we actually drive adoption,
which is a key part of winning that competition.
And so I do think some of the things that are being raised around how we make sure
that whatever we do doesn't actually only benefit the big companies that are already there,
but still allows new entrants, that we're protecting open weight models.
Those are all legitimate things that we need to be taking into account,
which is why we're saying we need to stay here and do the work and get it right
so that we can make sure that we're addressing these risks without, you know,
unintended consequences that might actually make things worse in the long term.
Yes, sir.
One of the options mentioned for the short term, if there's a way, was to have a gang of experts here.
You don't need permission to do that, do you?
And is that the plan?
No, we certainly don't need permission to do it,
but it would be nice if the Republican leadership would sort of convene and prioritize, right?
Right now the message and the signal going out to the entire Congress and the country
is that the Congress doesn't know enough about this technology.
We trust the executives of these companies to put guardrails in place themselves.
That is a far drive from we are going to put a working group together
to prioritize putting guardrails on technology to prevent catastrophic risk.
So, look, I have five children.
I would love to stay here, but I'm not going to waste my time here working by putting pen to paper
or striking a deal if the speaker is not actually sending a signal or a message to his own conference
that this is important and we are going to take it up as soon as we hit a level of consensus
that we can debate it on the floor.
You know, we had a bipartisan task force on AI that Vice Chairman Liu led,
and we had a number of really good recommendations that came out of it
and a number of bipartisan proposals that were ready to go, that would have passed on suspension,
that would have been a good first step,
and the Republican leadership refused to bring those bills up for a vote.
So, yes, we can stay on our feet.
I know we're all committed to keeping the conversations going,
but without a real commitment that there will be something that actually comes up for a vote,
it's hard to really drive towards the real consensus that we need.
Commissioner, recommendations?
Thank you for that question.
So we're working with Leo Jeffries' office that is in process.
We're still taking input from stakeholders.
So just yesterday the commission had a briefing on youth safety.
We had stakeholders come in, including two parents who were in chatbots.
It was pretty devastating to sit there and listen to what happened,
and I cannot imagine being a parent and having that.
I just want to go back to arguments that people make about China.
So, look, if chatbots in the U.S. are helping kids commit suicide,
I don't give a ---- what China does in regulating their chatbots
on letting their chatbots commit or help kids in China commit suicide.
We just need to do the right thing and make sure our chatbots here don't do that.
Same now as with cars, right?
If China doesn't put seatbelts on their cars or advance airbags on their cars,
we just need to do the right thing.
When would you like to put out those recommendations, like ideally?
That's a better question for Leo Jeffries.
Well, thank you very much for coming this morning,
and Speaker Johnson, with our bipartisan support,
trying to take this up again as soon as possible.
Thank you very much.
