There has been a flurry of activity at the Federal policymaking and regulatory level in recent weeks concerning artificial intelligence (AI) and technology in general, which seeks to keep up with the accelerating advancements and spending of industry and investors. Both the Legislative and Executive Branches are reacting to and seeking to address security in frontier models (e.g., Anthropic), safety mechanisms for children, consumers’ data and privacy security, protection against explicit material and copyright infringements, voter backlash to buildout job displacement and energy use, the financial leverage and windfall of mega firms, and do it all through an overarching regulatory regime acting as a preemptive Federal umbrella. Both branches must balance a prescriptive policy solution with the existential race for AI supremacy against China that is both an economic and national security imperative. Below is an update on some of the recent developments in Washington.
House Update
On Monday, the House Energy and Commerce Committee (E&C) announced a bipartisan agreement on kids’ online safety legislation (updated text here), reached by Chairman Brett Guthrie (R-KY) and Ranking Member Frank Pallone (D-NJ). The bill represents a unified House position, the Kids Internet and Digital Safety (KIDS) Act, consisting of language from various House bills, including: the SCREEN Act, the Kids Online Safety Act (KOSA), the Safe Messaging for Kids Act, the SPY KIDS Act, the Safer GAMING Act, the SAFE Bots Act, the Children and Teens’ Online Privacy Protection Act (COPPA 2.0), Data Broker Disclosures, the Safe Social Media Act, the No Fentanyl on Social Media Act, the Assessing Safety Tools for Parents and Minors Act, the Promoting a Safe Internet for Minors Act, the AWARE Act, the Kids Internet and Safety Partnership Act.
The revised package mandates that developers include age verification for sexually explicit websites and requires the platforms and their apps to curb harm to young people by using maximum default safety settings. In addition, the bill requires websites to seek parental consent to collect data from kids under 14 (up from 13). Key to the deal is language that sets a Federal policy floor that would supersede State laws that don’t meet its requirements for protecting minors while allowing States to enact laws that are stronger than the Federal standard. Unlike the Senate’s effort, however, the House package does not include a duty to “exercise reasonable care” in designing social media services to protect kids and teens. The bipartisan E&C product does carry House Republican and Democratic Leadership support and could move under suspension as soon as next week (notwithstanding the current Floor stalemate).
Finally, the E&C Energy Subcommittee approved the Ratepayer Protection Act (H.R. 9340) by voice vote yesterday, a bill that would codify President Trump’s initiative with industry. It is unclear how House Democrats will approach the bill when it hits the House Floor as the majority of the House Democratic Caucus has moved increasingly towards a full moratorium on new datacenter construction.
Senate Update
Meanwhile, senators have been racing to introduce legislation ahead of a July markup of AI legislation that Senate Commerce Committee Chairman Ted Cruz (R-TX) is planning. Chairman Cruz’s bill, introduced in April, would likely serve as a catch-all and includes the support of key Democratic senators like Adam Schiff (D-CA) and Brian Schatz (D-HI).
Senator Marsha Blackburn (R-TN) is hoping to reach an agreement with the White House by July 4 on a federal AI framework and KOSA bill with the hopes of being included in the Commerce markup. For some of the reasons outlined above, the House and Senate have differed in their approaches to kids’ safety, creating historical divisions between efforts in the two chambers. For now, the Administration appears to be backing the Blackburn-led effort, as it will likely include the most aggressive state preemption language of all the bills (which also makes Democratic support more difficult).
On Tuesday, a bipartisan group of senators introduced the SAFE KIDS Act, to regulate AI chatbots. Co-authored by Senators John Curtis (R-UT) and Adam Schiff (D-CA), the legislation would establish a comprehensive Federal framework for chatbot apps, requiring them to undergo “rigorous” risk assessments before being released publicly. The bill would also ban advertising to kids, mandate parental notifications and child safety audits, prohibit sexual deepfakes, and require age verification technology.
Separately, the Senate Judiciary Committee is planning to invite tech executives to testify at a July 28 hearing on safety and tech regulation. Executives the committee is considering inviting include Instagram’s Adam Mosseri, Snapchat’s Evan Spiegel, TikTok’s Adam Presser, and YouTube’s Neal Mohan.
Trump Administration
In addition to working with Senator Blackburn on an AI regulatory framework, the Administration has been active in its approach to security risks posed by AI models. The newfound focus is an evolution of sorts for President Trump. Between the Department of Commerce’s imposition of export controls and the Pentagon’s designation of Anthropic as a supply chain risk, the company has been under increased scrutiny by the Administration this year. In a recent interview with Axios, President Trump opined on using the Defense Production Act (DPA) to rein in the lab if needed, saying: “I have the power to use a lot of things. But I’m not sure I have to do that. It was a competitor [Amazon] and a part owner that turned Anthropic in. They didn’t like what they were doing. They were very concerned.”
Trump said that for now, the Administration is working collaboratively with Anthropic on standards to evaluate AI jailbreaks. And while the President has taken a more hawkish view on national security concerns surrounding the technology in recent months, he still ultimately sees the benefit of beating China in the technology race as outweighing the potential public harms.
Highlighting the dueling focus of the Administration, the President signed two Executive Orders (EO) on Monday to spur innovation in quantum information science and technology (QIST) while reaffirming national and data security in the digital space. The quantum order directs the Assistant to the President for Science and Technology (APST) to update the National Quantum Strategy and develop the first-ever quantum computer. The cyber order directs the Office of Management and Budget (OMB) and the National Cyber Director to lead a government-wide migration to post-quantum cryptography (PQC).
Meanwhile, the White House is getting outside help from FAI Action, the 501(c)(4) affiliated with the Foundation for American Innovation (FAI), according to Politico. FAI, which contributed to “Project 2025,” includes Senior Fellow Dean Ball, who is joining OpenAI (and was formerly at OSTP), and counts Michael Kratsios as a former board member. The group’s policy recommendations have been reflected in some of the Department of Government Efficiency’s (DOGE) early initiatives and in OMB’s overall approach to AI. One such recommendation was to reorient the U.S. Artificial Intelligence Safety Institute’s mission from ensuring AI safety to enabling “human flourishing and economic competitiveness.”
AI Buildout
Datacenters used 4–5% of the nation’s electricity in 2024, a share that is expected to triple by 2028, and some analysts predict nationwide electricity use will rise by as much as 20% over the next decade. Columbia University economist Stijn Van Nieuwerburgh estimates that AI buildout spending could top $8 trillion by 2032. With so much demand come rising prices for the inputs essential to the buildout, causing downstream pressures for the broader economy, particularly consumer-electronics. If the AI revolution proves as dramatic as many economists predict, those prices will eventually come down and have the potential to suppress long-term inflation. But those benefits are well into the future.
This creates an imperative for greater energy production as well as increased concerns over the costs of getting there. Those costs are seen through actual (or perceived) rate increases, job displacements, and environmental impacts. Public opinion sees perceived costs as actual—turning increasingly negative in its attitudes towards the AI buildout. Recent polling from Milltown Partners surveyed 6,872 registered voters in May, oversampling states with current datacenter projects (CA, GA, MI, NC, TX). While direct opposition is not yet a majority view, it continues to grow as nearly half of respondents support a temporary construction ban. Other notable findings:
Voters support datacenters being built near their homes, but the margin has shrunk to 38% in support and 34% in opposition.
A near majority (49%) support a moratorium on construction of new datacenters. Only 16% oppose a moratorium and 27% have not yet formed an opinion, meaning there is still potential for widespread movement in public opinion.
8% of the respondents who oppose datacenters say they know of one or more datacenters near their home. Translation: only a small fraction of datacenter opponents actually live near one.
The sentiment is driven by both the left and right, as Senator Bernie Sanders (I-VT) and Steve Bannon have attacked the technology as a threat to working people. Humans First, a conservative organization, is planning a “Nationwide Day of Protest” against the “unchecked expansion of AI data centers” on July 18.
Seeking to blunt the criticism of datacenters and find new sources of power to offset the costs and keep up with demand, the Department of Energy (DOE) announced it is providing $17.5 billion in loans to speed the development of 10 new large nuclear reactors. Secretary of Energy Chris Wright cited “tremendous interest” among developers that would buy the power, as well as utilities and energy companies. The nuclear plants could begin construction by 2030 and become operational within a decade. President Trump previously set a goal of quadrupling domestic production of nuclear power within the next 25 years. The Administration is also working to advance small modular nuclear reactors (SMRs).
Only two new large reactors have been built in the last three decades; both were billions over budget and delivered late. The 10 new proposed reactors would all use the same design, Westinghouse’s AP1000. Seven utilities and energy companies signed letters of intent that identified sites, of which DOE will pick five — hosting two reactors at each site. The utilities and Westinghouse are expected to contribute up to $5 billion in equity across the five sites. Wright said DOE will provide up to $17.5 billion in debt to pair with the equity.
Nvidia is also answering the concerns (many of which have been self-inflicted by the sector) by seeking to address the water and energy costs. The company recently unveiled a new cooling system that it says can dramatically reduce the amount of water and energy needed. The employment concern also shows signs of skewing in datacenters’ favor. Datacenters now face a labor shortage that companies are actively seeking to address, as evidenced by Meta’s recent $150 million skilled-labor investment. Zhou Xian, co-founder and CEO of Genesis AI, recently told Axios that his company is also focused on the labor shortage, saying “That’s the gap we want to fill.”
All of which is leading the Pentagon to lobby Congress against restricting its ability to build out datacenters on military facilities in the FY27 National Defense Authorization Act (NDAA). Emil Michael, the Pentagon’s top research and engineering official, has been pitching lawmakers on the growing national security imperative of datacenters to the Department of War’s mission.
OUTLOOK/ANALYSIS. The race for AI dominance and resulting policymaking environment have gained sizable new variables in just the last six months, spurred on by ground-breaking deployments of new frontier models, progress by Chinese labs in quickly closing their once-sizable gap with American competitors, and the various harm and displacement narratives of the AI buildout. All has converged in rising public opposition that Congress is particularly sensitive to, leading to a surge in campaign ads. Democrats’ position against the buildout has hardened and Republicans are now more willing to regulate where they once sided with a light government touch. While the preferred goal of the Administration is to allow U.S. developers the space to out-innovate, security concerns have added a nuance to its approach.
Ultimately, the safety concerns around America’s youth will have to act as the driving force for any AI policymaking to become a lawmaking exercise as it is the one issue set within the tech ecosystem that carries bipartisanship. How much additional weight kids’ safety can carry remains to be seen (particularly preemption), not to mention whether the two chambers have the ability to overcome growing animosity towards each other as the clock winds down on the current Congress. For now, the Administration is backing the Cruz/Blackburn effort, but the House is moving quickly to be a first mover and has already shown itself adept at avoiding the historical trend of conceding the Senate’s position throughout this Congress (e.g., reconciliation, housing, digital assets).
###