Artificial Intelligence

When AI is deployed in a workplace, it is almost always deployed against the workers, not for them. AI screens resumes and discards qualified applicants based on patterns no one is allowed to audit. AI sets schedules, monitors productivity, ranks employees against each other, and recommends who gets fired. The worker has no notice, no recourse, and no right to inspect the system that decides whether they keep their job.

We need mandatory disclosure when AI is used in hiring, firing, scheduling, evaluation, or compensation decisions. Workers should have the right to see what data was used about them and how the decision was reached. We need a federal floor on worker rights to notice and severance when AI is deployed at scale, and we need protected union bargaining over how AI gets deployed in workplaces.

The work cannot be done well if the people doing it are surveilled and ranked by black boxes designed to extract maximum output for minimum cost.

We need to be honest. Some jobs are going to be permanently transformed or eliminated by AI, not because AI is better at them, but because it is cheaper for the companies that own the AI. The history of automation is that the productivity gains accrue almost entirely to capital, not labor. We have decades of evidence on this.

Federal AI policy has to start from that fact. Workers being displaced did not cause the displacement. They should not bear the cost of it.

We need significantly expanded federal funding for AI displacement retraining, modeled on but bigger than the Trade Adjustment Assistance program. We need portable benefits that follow workers across jobs, not employers. We need a serious conversation about how to share productivity gains from AI deployment with the people whose work was used to train these models in the first place. And we need antitrust enforcement to prevent any single company from owning the foundational infrastructure of the next economy.

Federal agencies are deploying AI in benefits eligibility, fraud detection, immigration enforcement, criminal sentencing recommendations, and dozens of other contexts where the consequences of a mistake fall on individual citizens with no resources to fight back. The systems are often procured from private vendors, deployed without public review, and protected from scrutiny by trade-secret claims.

This is unacceptable. Any AI used by the federal government to make consequential decisions about citizens should be subject to mandatory algorithmic impact assessment before deployment, auditable by an independent body, documented in plain language so the affected person can understand how the decision was reached, and reversible through a human review process that is actually staffed and resourced.

I am familiar with the NIST AI Risk Management Framework. The framework exists. The political will to enforce it does not. That is the gap I will work to close.

AI-generated political content is here, and the legal framework has not caught up. We have already seen deepfake robocalls impersonating candidates, fabricated images deployed in political ads, and synthetic voice clones used to defraud and confuse voters.

We need mandatory disclosure of AI-generated content in political advertising, with real penalties for violations. We need a clear ban on AI-generated impersonation of candidates without consent. We need federal protections for election workers and officials targeted by AI-generated harassment campaigns, and transparency requirements for the platforms that distribute political content.

The First Amendment protects speech. It does not protect fraud.

A small number of companies own the foundational infrastructure of the AI economy. They control the chips, the data, the models, the cloud capacity, and increasingly the applications built on top of all of it. The same companies are also among the largest political donors in the technology sector and among the largest in any sector.

We have been here before. The country let railroads consolidate. The country let Standard Oil consolidate. The country let banks consolidate. Every time the consequences were the same: a small number of people got extraordinarily rich, and the rest of us had to fight for decades to claw back basic protections.

We need aggressive antitrust enforcement against AI consolidation, including review of cloud-AI tying arrangements that lock customers into a single vendor for both their compute and their models. We need restrictions on common ownership across the chip, model, and application layers. And we need a serious public conversation about whether foundational AI infrastructure should be a regulated utility or remain entirely private.

The people who know the most about AI risk are the people who build these systems. Many of them are increasingly alarmed by what they see inside these companies. They need legal protection to speak.

I have called out fraud at a major defense contractor and lost my career as a result. I know what it costs to tell the truth when the institution does not want to hear it. The story I have been hearing from current and former AI company employees is the same story, in a different industry, with worse non-disclosure agreements.

We need strong federal whistleblower protections for AI safety researchers, AI engineers, and employees of AI companies. We need mandatory safety incident reporting to a federal regulator with real enforcement authority. We need public-interest exemptions to the non-disclosure agreements that AI companies are using to silence employees from discussing legitimate safety concerns. And we need a funded, independent body to investigate AI safety failures with subpoena authority.

If a chemical plant has to disclose accidents to the EPA, an AI system being deployed at societal scale has to disclose its accidents too.