NEW:AI & the grid — 19-plank platformNEW:MRA spending dashboard
Research Brief
Published May 2026

The AI industry isn’t asking Congress to govern it. It’s asking Congress to make sure no one else can.

California is barely keeping up. State AI and data-center bills have been vetoed, weakened, and stripped before passage. Now the industry is pushing federal preemption to override even that.

I'm a climate scientist running for Congress in California’s 32nd congressional district (CA-32). This page documents what we know — from primary sources — about the convergence of artificial intelligence, cryptocurrency, and electricity infrastructure, what it would mean for our district if state authority is preempted, and what a federal policy response should look like.

176 TWh
Terawatt-hours of U.S. data center electricity consumption in 2023 — roughly 8× the entire residential electricity supply of all 10 million LA County residents. Projected to grow to 15–28× LA County residential by 2028.[1]
$1.1T
Projected investor-owned utility capital expenditures, 2025–2029. Ratepayers fund this through their bills.[2]
10 yrs
Length of the federal moratorium on state AI laws proposed in 2025. Stripped from the One Big Beautiful Bill (OBBB) before passage. Expected to return.[16]
18.7 GW
Gigawatts of data center service capacity requested in California: more than enough to power every household in the state.[8]
Section 01

The fight is preemption.

The AI industry’s strategic goal in Washington is not to pass federal AI law. It is to prevent California, New York, Colorado, and every other state from passing their own.

In 2025, an industry-backed ten-year federal moratorium on state and local AI laws was attached as a rider to the omnibus budget package. It would have nullified California's Senate Bill 53 (SB 53), New York's Responsible AI Safety and Education (RAISE) Act, and every other state framework currently on the books. The Senate stripped it before passage. Industry lobbyists have made clear it will return.

At the same time, the industry is funding direct attacks on state-level legislators who have written enforceable AI safety law. Leading the Future, a political action committee (PAC) — backed by Andreessen Horowitz, Joe Lonsdale, Perplexity, Ron Conway, and others — and its affiliate Think Big PAC have spent more than $1.5 million in independent expenditures against Alex Bores, the New York assemblyman who co-authored the RAISE Act, in his congressional bid alone. NOTUS, a national political news outlet, has reported the strategy bluntly: the goal is to beat up on Bores so badly that other politicians run the other direction whenever AI regulation comes up.

The pattern is consistent: (1) defeat strong bills at the state level through gubernatorial vetoes. (2) Weaken what survives through legislative bargaining. (3) Override what remains through federal preemption. These three strategies are being executed simultaneously.

Section 02

What California has — and what it has lost.

The California state record on AI and data-center governance is not a model of regulatory leadership. It's a holding action that has already been substantially defeated. The bills that passed are weaker than the bills that were introduced, and the bills that were vetoed were the strongest ones. The pattern below shows the limit of state authority even before federal preemption is legislated (e.g. the 10-year AI regulation moratorium that was stripped from the OBBB).

SB 1047 (Wiener, 2024)
VETOED

The strongest AI safety bill in the country: full shutdown capabilities for frontier models, third-party audited risk assessments, downstream developer liability, whistleblower protections, and a public CalCompute cluster. Newsom vetoed it September 29, 2024 after a sustained lobbying campaign by Andreessen Horowitz, OpenAI, and Meta.

SB 53 (Wiener, 2025) — Frontier AI Transparency Act
WEAKENED

Signed September 2025 as the streamlined successor to SB 1047. Covers developers training models on more than 10²⁶ floating-point operations of compute (the same scale as the federal AI Executive Order), with stricter disclosure and incident-reporting obligations triggered when annual revenue exceeds $500 million. New York's RAISE Act uses the same 10²⁶ compute floor but adds a parallel trigger — any developer that has spent more than $100 million on training-compute costs — that catches some smaller bespoke developers California's revenue tier doesn't reach. The strongest pieces of the bill were stripped during the legislative process: third-party audits were dropped, and downstream developer liability was dropped. What survived: published safety frameworks and a 15-day critical-incident reporting requirement (24 hours for incidents posing imminent danger).

SB 57 (Padilla, 2025) — Ratepayer Protection
WEAKENED

Signed October 2025. As enacted, requires only a study by the California Public Utilities Commission (CPUC) by January 2027 of how data center demand shifts costs to other ratepayers. Earlier versions proposed an actual electricity tariff requiring data centers to offset the costs they shift to other consumers; the Assembly Appropriations Committee stripped that provision in late August 2025 under data center industry lobbying. CalMatters documented the strip in detail, quoting The Utility Reform Network: lobbyists for data centers “successfully gutted the bill.”

AB 93 (Bauer-Kahan, 2025) — Data Center Water Disclosure
VETOED

Would have required data center operators to disclose water use, including water embedded in cooling and electricity generation. Vetoed by Newsom October 2025. Without disclosure, neither regulators nor the public can determine the resource cost of data center buildouts.

SB 7 (Smallwood-Cuevas, 2025) — No Robo Bosses Act
VETOED
Would have required employer disclosure when using automated decision systems in employment decisions. Labor unions supported it; chambers of commerce opposed. Vetoed by Newsom October 2025.
AB 1064 (Bauer-Kahan, 2025) — Leading Ethical AI Development (LEAD) Act
VETOED
Would have prohibited companion chatbots from harmful interactions with minors. Vetoed by Newsom October 13, 2025; his veto message argued the bill risked an effective ban on minors' AI use rather than letting them “learn how to safely interact with AI systems,” and that overlapping protections were already covered by SB 243, which he signed.
SB 11 (Ashby, 2025) — AI Consumer Warnings
VETOED
Would have required consumer warnings on AI digital replica tools and clarified that false impersonation law applies to deepfakes. Vetoed by Newsom October 2025.

The pattern is consistent. Strong bills are vetoed. Weakened bills get signed, and whatever survives is what federal preemption is now positioned to erase.

Section 03

What CA-32 stands to lose.

CA-32 voters do not need a theoretical argument about why corporations should not be trusted to police themselves. We've lived through it.

In 2015 and 2016, SoCalGas's Aliso Canyon storage facility released roughly 97,100 metric tons of methane over four months — the largest accidental natural gas leak in U.S. history. Eight thousand Porter Ranch households were displaced. Methane is roughly eighty times more potent than carbon dioxide as a greenhouse gas over a twenty-year horizon. SoCalGas was permitted to continue operating the facility. Health complaints from displaced residents continue to this day.

The Santa Susana Field Laboratory — Boeing-owned, formerly operated by Rocketdyne — was the site of a partial nuclear reactor meltdown in 1959 and decades of subsequent rocket-engine and nuclear testing. Cleanup obligations among Boeing, NASA, and the Department of Energy have been disputed and litigated for decades. Cleanup deadlines have been pushed back repeatedly. The contamination still has not been remediated — nearly sixty-seven years after the meltdown.

In November 2018, the Woolsey Fire burned across the still-contaminated SSFL site. A peer-reviewed study in the Journal of Environmental Radioactivity in 2021 found that fire ash carried SSFL-related radioactive microparticles to neighborhoods around the burn perimeter. Post-fire debris from the affected area was hauled to LA County landfills, including the Calabasas Landfill, on the boundary of CA-32. Boeing’s failure to remediate became Calabasas’s burden to absorb. Meanwhile in Castaic, just north of the district, the Chiquita Canyon Landfill has been running an underground heat-releasing chemical reaction since early 2022. By 2024 residents had filed more than 9,400 odor complaints; LA County sued the operator; the Department of Toxic Substances Control issued an “imminent and substantial endangerment” order. The landfill ceased operations in December 2024, but its underground reaction has continued to grow.

None of these were data center failures. They are the pattern of corporate environmental impunity that the data center industry is now being inserted into. And the geographic lesson is the same in every case: the harm doesn’t have to originate inside CA-32 to land here. Boeing’s contamination didn’t stop at the district line, because the fires disturbed it and the winds distributed it. And the waste was transported to Calabasas. Aliso Canyon’s methane plume didn’t ask permission to settle over Porter Ranch, and the other contaminants are believed to have entered our groundwater. The pollution from Chiquita Canyon’s unremitting fires drifted on the prevailing winds. The data center buildout will follow the same pattern: most facilities will be sited east and inland of us, but the ratepayer cost, the wildfire-hardening trade-off, the air pollution from diesel backup generators, and the landfill burden of obsolete servers will all flow downstream to where we live.

Loading district map…
CA-32 (blue) sits inside the service territory of Southern California Edison (SCE) and the Los Angeles Department of Water and Power (LADWP), which extends east into the Inland Empire and Antelope Valley, where most new data centers are sited. The marked sites — Calabasas Landfill, Chiquita Canyon Landfill, the Santa Susana Field Laboratory (SSFL), Aliso Canyon — show recent histories of corporate environmental impunity inside or immediately adjacent to the district. CA-32 boundary from California Assembly Bill 604 (AB 604, Prop 50) redistricting; site locations are approximate.

CA-32 does not host hyperscaler data centers. That is the point.

The question is not whether data centers are coming to Northridge or Calabasas; the question is whether CA-32 ratepayers will subsidize the data center buildout happening across Southern California Edison and LADWP service territory — while wildfire hardening, undergrounding, and distribution upgrades in their own neighborhoods compete for the same capital.

That allocation decision is currently made by California state regulators, but federal preemption is the move to take it away.

The state-level tools that determine the answer

Every one of the tools below is currently held at the state level. Each is a target for federal preemption.

CPUC cost-allocation rules
What it does
Decides whether new transmission and generation built to serve data centers gets paid for by all ratepayers, or only by the data center operators causing the cost.
What preemption would erase
A federal preemption of state utility cost-allocation — through a Clarity-Act-adjacent rider, or a deregulatory rule reset by the Federal Energy Regulatory Commission (FERC) — would lock in residential cross-subsidization and remove California's authority to redirect capital to wildfire hardening.
CEC reporting (AB 222 framework)
What it does
Forces data centers to disclose energy use, water use, Power Usage Effectiveness, and the share of energy going to AI training. Necessary input for any informed policy.
What preemption would erase
A weak federal disclosure floor would override California's stronger one. The pattern: state minimum becomes state ceiling when federalized.
CEQA environmental review
What it does
The California Environmental Quality Act (CEQA) triggers analysis of grid, water, and air-quality impacts before approval. Gives communities legal standing to challenge inadequate review.
What preemption would erase
Federal ‘permitting reform’ proposals have repeatedly capped state environmental review for ‘critical infrastructure.’ Data centers would be classified as such.
CARB emissions oversight
What it does
Diesel backup generators at data centers are a significant source of fine particulate matter (PM2.5) air pollution. The California Air Resources Board (CARB) has been moving to limit them.
What preemption would erase
Federal preemption of state air-quality rules for ‘energy-critical infrastructure’ would remove this lever.
Local zoning and permitting
What it does
Counties and cities can refuse or condition large industrial loads.
What preemption would erase
Already partially preempted (wireless siting precedent). A federal data-center permitting framework would extend that pattern to AI infrastructure.
State legislature
What it does
Each new bill is a tool added to the kit. SB 53, SB 57, AB 222, future legislation.
What preemption would erase
A 10-year federal moratorium on state AI laws would freeze state authority for a decade. The 2025 attempt was struck before passage. It will return.
The through-line

Every dollar of SCE or LADWP capital that flows into transmission serving inland data centers is a dollar not flowing into wildfire hardening here. Today, California has the authority to require that data center operators pay their own cost. Federal preemption is the move to take that authority away. That is what is at stake for CA-32 voters in this fight, even though the data centers themselves are being built somewhere else.

But communities have power. Monterey Park just proved it.

Through a series of unanimous Council actions across early 2026 — a 45-day moratorium adopted in January, an extension of that moratorium in March (running through January 2027), and ordinances in April formalizing a permanent prohibition — the Los Angeles County city of Monterey Park became the first city in California to ban data-center construction within its limits. The actions followed months of community organizing that pressured an Australian developer (HMC Capital StratCap) to withdraw a 250,000-square-foot proposal at 1977 Saturn Street. Voters will confirm the permanent ban via Measure NDC (“No Data Center”) on the June 2 ballot.

This is what local authority looks like when the federal government doesn’t take it away. It is exactly the kind of self-organizing common-resource governance that Elinor Ostrom documented — the work that won her the 2009 Nobel Prize in Economics. In her 1965 UCLA dissertation on the West Basin groundwater case (later incorporated into her 1990 book Governing the Commons), Ostrom showed that local users of the West Basin aquifer in LA County resolved a saltwater-intrusion crisis without privatization and without top-down regulation: they organized the West Basin Water Association, set pumping limits among themselves, and reduced groundwater withdrawals by approximately thirty percent over two decades. The federal government did not preempt their authority to do it.

A federal moratorium on state and local AI laws would not just override Sacramento. It would override Monterey Park.

Section 04

The evidence.

The scale of the buildout, the cost shift to ratepayers, and the grid consequences all matter to the preemption fight because they explain why the industry needs preemption in the first place: governing this honestly — under any rate-allocation rule that makes hyperscalers pay for what they cause — would slow the buildout. Preemption removes the constraint.

U.S. data center electricity consumption — actual and projected

Terawatt-hours (TWh) per year. The solid black line is historical Lawrence Berkeley National Laboratory (LBNL) actuals through 2023. The dashed red lines are LBNL’s 2028 projection range (low and high scenarios) — estimated projections, not measured data.[1]

Historical (LBNL actuals, 2014–2023)
Estimated 2028 projection (LBNL low & high scenarios)
Range of uncertainty between projections
01282563845126402014201820232028Estimated projection — low: 176 (2023)Estimated projection — low: 325 (2028)Estimated projection — high: 176 (2023)Estimated projection — high: 580 (2028)Historical (TWh): 58 (2014)Historical (TWh): 76 (2018)Historical (TWh): 176 (2023)325580TWh per year
Hover over a data point to see details
Estimated projection — low
Estimated projection — high
Historical (TWh)

The 2028 projection range (325–580 TWh) corresponds to 6.7 to 12 percent of total U.S. electricity demand. Two ways to put that in perspective:

  • Local scale. U.S. data centers in 2023 already consumed roughly 8× the entire residential electricity supply of all 10 million LA County residents. The 2028 projection range is roughly 15× to 28× LA County residential. (Calculation: ~3.4M LA County households (per US Census American Community Survey, ACS) × ~6,200 kWh/yr (per US Energy Information Administration (EIA) Residential Energy Consumption Survey, Pacific region) ≈ 21 TWh/yr LA County residential total.)
  • Climate opportunity cost. The same 176 TWh used by data centers in 2023 would be enough to electrify roughly one in six U.S. passenger vehicles — about 50 million cars off internal combustion. The 2028 high projection (580 TWh) is enough to electrify more than half the entire U.S. passenger fleet. (Calculation: ~3 trillion U.S. vehicle-miles per year (per Federal Highway Administration (FHWA) Traffic Volume Trends; light-duty vehicles are the dominant component) ÷ ~3 mi/kWh average electric vehicle (EV) efficiency ≈ ~1,000 TWh/yr to fully electrify the U.S. passenger fleet.)
“Data center load growth has tripled over the past decade and is projected to double or triple by 2028.”
— Lawrence Berkeley National Laboratory, 2024

Investor-owned utility capital expenditures, 2020–2025

Billions of U.S. dollars per year. The Edison Electric Institute (EEI) projects $1.1 trillion in capital expenditures from 2025 through 2029, nearly matching the combined $1.3 trillion spent over the entire prior decade.[2]

$0B$60B$120B$180B$240B$300B2020: $139.8B$139.8B20202021: $145B$145B20212022: $158B$158B20222023: $168B$168B20232024: $178.2B$178.2B20242025: $207.9B$207.9B2025$ Billions
2020: $139.8B
2021: $145B
2022: $158B
2023: $168B
2024: $178.2B
2025: $207.9B

Who is paying for this?

When a data center connects to a regional grid, the utility builds new transmission, substations, and generation to serve it. Under most current rate structures, those costs get spread across all ratepayers — including residential customers who never agreed to subsidize hyperscaler buildouts.

In the Mid-Atlantic regional grid (PJM Interconnection, the wholesale electricity market serving 13 states plus DC, from New Jersey west to Illinois and south to North Carolina), the independent market monitor reported in October 2025 that data centers were “the primary reason” for the 82 percent surge in capacity auction revenue, which jumped from $8.8 billion to $16.1 billion in a single year.

That cost is now flowing to consumer bills across thirteen states. Carnegie Mellon researchers project that under current policy, electricity rates in Northern Virginia — the largest concentration of data centers in the country — will rise more than 25 percent by 2030, with more than 25 gigawatts of aging coal plants kept online to meet data center demand.

PJM auction revenue
$8.8B → $16.1B
Single-year jump in PJM grid capacity auction revenue. The independent market monitor attributed the increase primarily to data center demand.[3]
Section 05

A Green New Deal for the compute era.

The original Green New Deal framework was authored in 2019, before ChatGPT, before the data center buildout, before crypto’s full grid footprint became visible.

Its core principles still hold: a just transition, a hardened grid, public infrastructure as a public good, climate science as the guidepost. What follows is an updated platform — nineteen federal policy planks organized into five groups: the federalism fight (planks 1–3), frontier AI safety (planks 4–8), data centers and the grid (planks 9–13), cryptocurrency (planks 14–15), and workers, kids, and consumers (planks 16–19). Plank 1 is the headline. The rest of the platform derives from it.

Group 01

The federalism fight.

Who gets to govern AI, data centers, and consumer technology — Washington, or the states and cities directly affected? Plank 1 prevents the harm; Plank 2 enforces against rollback; Plank 3 names the positive principle.

01

No federal preemption of state AI safety, data-center, or consumer-protection laws

This is the single most important AI policy battle in the next Congress. I will vote against any preemption of state AI safety, transparency, or consumer protection laws — whether as standalone legislation, as riders attached to broader bills, as multi-year moratoria, or as conditions on federal funding that require states to give up their authority.

This includes preemption of state utility cost-allocation rules, state data-center reporting requirements, state environmental review, and state air-quality oversight. Each one is a tool that determines whether the cost of the AI buildout is borne by hyperscalers or by ordinary ratepayers in places like CA-32.

02

Defense of state action against federal overreach

Public support for state attorneys general (AGs) bringing actions to defend their AI laws against industry challenges. Amicus briefs as needed. Use of the appropriations process to deny federal agencies funding to undermine state safety frameworks.
03

Federal recognition of community-led commons governance

Federal AI and data-center policy must explicitly recognize and protect the authority of cities, counties, and tribal governments to regulate facilities within their jurisdictions — including the authority to impose moratoria, ban specific land uses, set water and energy disclosure requirements, and condition permitting on community impact review. Monterey Park's April 2026 ban on data-center construction is a direct exercise of this authority and the kind of locally-grounded governance that any federal framework should preserve, not preempt.

The intellectual foundation here is well-established. Elinor Ostrom's Nobel-recognized research on governing the commons — including her doctoral work on Los Angeles groundwater — demonstrated that shared resources are most sustainably managed through nested institutions that include real local control, monitoring, and enforcement. Federal preemption is the opposite move, as it dismantles the nested structure that makes local governance work. Plank 1 prevents the harm. Plank 3 names what we are protecting.

Group 02

Frontier AI safety.

Direct regulation of the most capable models: a federal floor based on the RAISE Act, whistleblower protections, strict liability for catastrophic harms, training-data transparency, and a public alternative to industry concentration.

04

Federal AI safety floor based on the RAISE Act

A federal AI safety law should set a floor that states can exceed, not a ceiling that overrides them. The RAISE Act framework — third-party safety audits, 72-hour critical incident reporting, dedicated oversight office funded by fees on developers, AG civil enforcement — should be the federal model.
05

Whistleblower protections for AI lab employees

Codified federal protections for employees of AI labs, graphics processing unit (GPU) manufacturers, and infrastructure providers who report safety concerns. Voluntary commitments are unenforceable. Whistleblowers are the only mechanism that has consistently surfaced industry practices.
06

Strict liability for catastrophic AI harms

Rather than legislators specifying every safety practice in advance, presume developers liable for harms caused by their models unless they can demonstrate reasonable care. This aligns economic incentives toward safety without micromanaging fast-moving technology.
07

Federal training data transparency

A federal version of California SB 942 and New York AB 6578: AI developers must publicly disclose training data sources, including third-party data acquisitions. Necessary for copyright accountability, scientific reproducibility, and public oversight. The framework should also extend the principles negotiated by the Screen Actors Guild–American Federation of Television and Radio Artists (SAG-AFTRA) in their 2023 contracts — informed consent and compensation for the use of an individual's name, image, voice, or likeness in AI training and output — to all creative workers federally, not just those represented by a union with leverage to negotiate.
08

Public compute infrastructure

A federal public cloud cluster — accessible to academic researchers, public-interest developers, small companies, and civil society — to prevent AI capability from being concentrated in five corporate hands. Funded through the Department of Energy, hosted at national labs and university partners. The CalCompute provision was stripped from California's SB 1047 before its veto. Congress should build it federally.
Group 03

Data centers and the grid.

The infrastructure economics. Who pays for the buildout, what gets reviewed before it's built, and how grid capital keeps flowing to wildfire hardening rather than to hyperscaler subsidies.

09

Data center cost allocation reform

Federal legislation directing FERC to require that large new electrical loads — data centers above a defined threshold — pay the full incremental cost of grid infrastructure required to serve them. The Texas SB 6 model adapted to interstate transmission. Residential ratepayers should not subsidize hyperscaler buildouts. This must be drafted as a floor that preserves state authority to require more, not a ceiling that overrides it.
10

Federal data center reporting standards

Mandatory federal reporting of data center energy consumption, water consumption, Power Usage Effectiveness metrics, and emissions to a centralized public database. Modeled on California AB 222 — and structured so that states with stronger reporting are not penalized.
11

Grid hardening and infrastructure prioritization

Federal infrastructure dollars should prioritize wildfire hardening, undergrounding in fire-prone districts, distribution-level resilience, distributed generation, and battery storage — and explicitly disallow capture by transmission projects whose primary beneficiary is hyperscaler data centers. The Inflation Reduction Act (IRA) and the Infrastructure Investment and Jobs Act (IIJA) both have authority that has not been used aggressively for this purpose.
12

Stranded asset risk protection — and no federal bailout for AI or crypto

If the AI data center boom moderates or reverses, ratepayers should not be left holding the bag for transmission and generation built to serve facilities that never operate. Federal rules should require data center operators to post bonds or pre-pay for grid infrastructure they cause to be built, with refunds only after sustained operation.

The same principle extends to the broader stability question: there should be no federal bailout for AI or cryptocurrency firms whose business models fail. Industry concentration risks must not be socialized. The too-big-to-fail framework that drove the 2008 financial crisis cannot be allowed to repeat with these technologies. AI stock market gains are widely believed to be a bubble; if it pops, the loss belongs to the firms and their investors, not to taxpayers.

13

National Environmental Policy Act (NEPA) review of major data center projects

Data center projects above defined energy or water thresholds should be subject to environmental review. Many projects currently evade NEPA by being sited on private land with private power purchase agreements, despite imposing enormous public infrastructure costs.
Group 04

Cryptocurrency.

A distinct industry with a distinct legislative fight. Mining standards parallel to data-center oversight, and substantive opposition to the Clarity Act.

14

Cryptocurrency mining standards

Federal disclosure requirements for cryptocurrency mining operations parallel to data center reporting. Restrictions on the use of fossil-fuel-fired generation primarily dedicated to mining loads. State authority to impose mining moratoria preserved against federal preemption challenges. New large-scale data centers and cryptocurrency mining facilities should be required to operate on 100% carbon-free energy generation — a commitment also reflected in my issues platform — so that the AI and crypto buildout does not lock in another generation of fossil-fuel demand.
15

Oppose the Clarity Act

The Digital Asset Market Clarity Act would strip the Securities and Exchange Commission (SEC) of meaningful jurisdiction over the cryptocurrency industry and shift oversight to the Commodity Futures Trading Commission (CFTC), which lacks the staffing and authority to regulate retail investor protection at scale. It is the cryptocurrency industry's number-one federal legislative priority. Rep. Brad Sherman and Rep. Maxine Waters led the opposition to its predecessor, the Financial Innovation and Technology for the 21st Century Act (FIT21), on the House Financial Services Committee.

I will vote against the Clarity Act. The reasons are substantive: it weakens consumer protection in a sector where fraud has cost ordinary investors billions; it entrenches the regulatory permissiveness that allows energy-intensive proof-of-work mining to externalize its grid costs; and it codifies the political conditions under which schemes like World Liberty Financial — the Trump-family crypto venture — can enrich political insiders without meaningful oversight.

The crypto industry has spent over $245 million through Fairshake and its affiliated super PACs in the 2024 cycle, including more than $40 million to defeat Sen. Sherrod Brown in Ohio — a senior senator who chaired the Banking Committee and questioned the industry's regulatory carveouts. The same industry is now active in the 2026 primary cycle. Every Democratic candidate in this race should answer, before primary day, whether they will commit to voting against the Clarity Act.

Group 05

Workers, kids, and consumers.

The human-facing protections. Updates to the Worker Adjustment and Retraining Notification (WARN) Act for AI-driven displacement, federal kids' safety on AI products, a national consumer data privacy law, and an AI Dividend that redistributes the upside.

16

Worker protections for AI-driven displacement

An updated WARN Act with explicit provisions for AI-driven workforce reductions: extended notification periods, retraining funding, severance floors. The framework of California SB 7 (vetoed by Newsom in 2025) — disclosure of automated decision systems used in employment — should be federal law.
17

Federal protections for kids on AI products

AI products designed for or accessible by minors require federal safeguards. I will support legislation requiring mandatory age verification for high-risk generative AI tools (image generators, companion chatbots, deepfake services), parental consent for AI products marketed to children under sixteen, chatbot screening for self-harm content, mandatory federal reporting of AI-related harm incidents involving minors, and federal criminal penalties for the production or distribution of AI-generated child sexual abuse material. The framework of California's vetoed LEAD Act (AB 1064, Bauer-Kahan) should be enacted federally and strengthened. Children must not be the testing ground on which the AI industry calibrates its liability exposure.

18

A national consumer data privacy law

The United States is one of the only major democracies without a comprehensive federal consumer data privacy law. I will support a national baseline that establishes individual ownership of personal data, requires affirmative consent for sale or transfer, guarantees meaningful access and deletion rights, and extends Plank 7's training-data disclosure requirement to all consumer-facing AI systems. Critically — and consistent with Plank 1 — the federal floor must explicitly preserve the authority of states like California to maintain stronger protections. As with AI safety, a federal floor cannot become a state ceiling.
19

An AI Dividend — tax frontier developers, share the surplus

This entire platform is built on the recognition that AI and cryptocurrency companies have spent the past decade privatizing the profits of new technology while socializing its costs — onto ratepayers funding grid expansion, onto communities absorbing landfill burden, onto workers displaced without recourse, onto residents breathing diesel exhaust from backup generators. Plank 19 inverts the equation directly.

The companies building artificial intelligence are doing it on the back of public infrastructure: federally-funded research that produced the foundational architectures, public-domain text and image data, decades of taxpayer-funded basic science, and the creative output of millions of people whose work was scraped without consent or compensation. The upside is being captured by five firms.

I will support an AI Dividend funded by a federal excise tax on the largest AI developers (revenue threshold set high enough to apply only to frontier-tier firms). Funds would support three purposes simultaneously: direct payments to American workers in industries most exposed to AI-driven displacement; investment in green jobs and community energy resilience in districts absorbing the data-center buildout’s grid impact; and capacity for displaced workers to transition into the public-interest sectors named in Plank 16. This is the offensive complement to Plank 16’s defensive worker-protection framework: not just cushion the downside — redistribute the upside. The privatize-profit, socialize-cost pattern of the past decade ends here.

Section 06

Glossary & citations.

Glossary of acronyms

Each acronym is defined inline on first use in the body of this brief. The list below is a single-spot reference for any reader who arrived mid-page or wants a quick refresher.

AB
Assembly Bill (California state legislature)
ACS
American Community Survey (US Census Bureau)
AG
Attorney General
CA-32
California's 32nd congressional district
CARB
California Air Resources Board
CEC
California Energy Commission
CEQA
California Environmental Quality Act
CFTC
Commodity Futures Trading Commission
CPUC
California Public Utilities Commission
EEI
Edison Electric Institute
EIA
US Energy Information Administration
EV
Electric vehicle
FERC
Federal Energy Regulatory Commission
FHWA
Federal Highway Administration
FIT21
Financial Innovation and Technology for the 21st Century Act
GPU
Graphics processing unit
GW
Gigawatts (1 GW = 1,000 MW = 1 billion watts)
IIJA
Infrastructure Investment and Jobs Act (2021)
IRA
Inflation Reduction Act (2022)
LADWP
Los Angeles Department of Water and Power
LBNL
Lawrence Berkeley National Laboratory
LEAD Act
Leading Ethical AI Development Act (California AB 1064, vetoed 2025)
NDC
“No Data Center” (Monterey Park ballot measure)
NEPA
National Environmental Policy Act
OBBB
One Big Beautiful Bill (2025 federal omnibus budget package)
PAC
Political action committee
PJM
PJM Interconnection — wholesale electricity market serving 13 states plus DC, from New Jersey west to Illinois and south to North Carolina
PM2.5
Fine particulate matter (airborne particles ≤ 2.5 micrometers)
RAISE Act
Responsible AI Safety and Education Act (New York)
SAG-AFTRA
Screen Actors Guild–American Federation of Television and Radio Artists
SB
Senate Bill (California state legislature)
SCE
Southern California Edison
SEC
Securities and Exchange Commission
SSFL
Santa Susana Field Laboratory
TWh
Terawatt-hours (1 TWh = 1 billion kilowatt-hours)
WARN Act
Worker Adjustment and Retraining Notification Act

Citations

  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
A note on methodology

Every numerical claim on this page traces to a primary source — federal agency reports, peer-reviewed research, official utility filings, or independent market monitors. I am a research scientist. I do not believe a campaign should hold itself to a lower evidentiary standard than a peer-reviewed paper. If you find an error, write to contact@marenalinforcongress.com and I will correct it.

Read the rest of the platform

This is one piece of a broader campaign platform on climate, healthcare, immigration, housing, and democracy reform.

See full platformGet involved
AI, the Grid, and the Fight Over Preemption — Marena Lin for Congress