Artificial Intelligence: Innovation Needs Guardrails, Not Just Speed

Why Does the US Have No AI Law?

The most powerful technology in a generation is developing with zero federal regulation. The United States has not passed a single comprehensive law governing artificial intelligence — not on bias, not on transparency, not on safety, not on deepfakes, not on anything. The reasons are structural, political, and deliberate.

Industry lobbying. The six companies that control over 90% of frontier AI development — Google, Microsoft, Meta, Amazon, OpenAI, and Anthropic — spend hundreds of millions annually on lobbying and have successfully framed regulation as a threat to American competitiveness. Their preferred approach is "self-regulation," which is to say, no regulation. This is the same playbook used by Big Tobacco, Big Oil, and Big Finance before them. In every case, self-regulation failed to prevent the harms that regulation was designed to address.

Congressional technological illiteracy. The average age of a US senator is 65. The average age of a House member is 58. Congress famously struggled to understand basic concepts during Facebook hearings and has shown no greater fluency with AI. Members lack the technical staff to draft sophisticated legislation, and the committee structure doesn't map to the cross-cutting nature of AI policy. The result: hearings that produce headlines but not bills.

The EU moved first. The European Union passed its AI Act in 2024 — a comprehensive, risk-based regulatory framework that classifies AI systems by potential harm and applies proportionate rules. The UK established its AI Safety Institute. Canada passed the Artificial Intelligence and Data Act. Meanwhile, California's attempt at state-level AI regulation was vetoed by the governor under industry pressure. The US is not leading on AI governance — it has ceded the field entirely.

The absence of law is itself a policy choice — one that benefits the companies building AI and harms everyone affected by it. Until Congress acts, Americans have no recourse when an algorithm denies them a job, a loan, bail, or medical care. Read more about the regulatory gap on our privacy and internet freedom page.

What Are the Real Risks of Unregulated AI?

The risks of AI without guardrails are not hypothetical — they are happening now. From biased hiring algorithms to election- threatening deepfakes, the consequences of unregulated AI are already measurable and growing.

• Deepfakes and Elections: AI-generated audio and video can now create nearly undetectable fabrications of political figures saying things they never said. In 2024, AI-generated robocalls impersonated President Biden to suppress voter turnout. With no federal deepfake law, this will escalate in every future election cycle.
• Algorithmic Bias: AI systems trained on historical data replicate and amplify human biases. Hiring algorithms discriminate against women. Credit scoring models penalize minority communities. Criminal sentencing tools assign higher risk scores to Black defendants. These systems make consequential decisions about people's lives with no transparency and no recourse.
• Job Displacement: McKinsey estimates 30% of work hours could be automated by 2030, with 300 million jobs globally at risk of significant transformation. The pace exceeds previous automation waves, and current US worker retraining programs — funded at 0.1% of GDP compared to Denmark's 2% — are nowhere near adequate.
• Privacy Erosion: AI enables surveillance at scale: facial recognition without warrants, predictive policing based on zip codes, consumer tracking across platforms. Without federal privacy law, there are no meaningful limits on what AI systems can collect, infer, and act upon regarding individual behavior.
• Autonomous Weapons: AI-powered weapons systems that can select and engage targets without human intervention are being developed by the US, China, and Russia. There are no international treaties governing autonomous weapons, and the potential for algorithmic error, escalation, and erosion of the laws of war is severe.
• Concentration of Power: Six companies control the compute, the data, and the talent for frontier AI development. This concentration gives a handful of corporations unprecedented influence over the infrastructure of information, communication, and decision-making — with no democratic accountability.

These risks are not arguments against AI — they are arguments for governing it. Every transformative technology in history — aviation, pharmaceuticals, nuclear energy, financial markets — required regulation to deliver its benefits without unacceptable harm. AI is no different. See our cybersecurity page for more on the digital threat landscape.

How Does the Common Good AI Plan Work?

The Common Good plan establishes a federal AI regulatory framework that promotes innovation while protecting Americans from the harms that unregulated AI is already causing. It is modeled on the EU AI Act's risk-based approach, adapted for American priorities.

The plan is built on eight core provisions, each targeting a specific gap in current governance. Together, they create a framework where AI development continues to thrive while its most dangerous applications are subject to democratic oversight.

• Federal AI Regulatory Framework: Establish a risk-based classification system for AI applications. Minimal-risk tools (spam filters, autocomplete) face no new regulation. High-risk tools (hiring, lending, criminal justice, healthcare) require transparency, auditing, and human oversight. Unacceptable-risk tools (mass social scoring, manipulative systems targeting children) are banned.
• Algorithmic Transparency Requirements: Companies deploying AI in high-stakes decisions must disclose that AI is being used, explain how it reaches decisions, and provide meaningful avenues for appeal. You have the right to know when an algorithm denied you a job, a loan, or bail — and the right to challenge that decision.
• Mandatory Bias Auditing: All AI systems used in hiring, lending, criminal justice, and healthcare must undergo independent bias audits before deployment and on a regular schedule thereafter. Results must be made public. Companies that deploy discriminatory AI systems face civil liability.
• Deepfake Labeling and Criminalization: Require AI-generated content to carry watermarks and labels. Criminalize the creation and distribution of deepfakes intended to defraud, manipulate elections, or create non-consensual intimate images. Establish a federal reporting mechanism for deepfake abuse.
• Worker Transition Programs: Go beyond retraining rhetoric. Establish wage insurance for displaced workers (covering 50% of lost wages for up to two years), portable benefits not tied to employers, expanded community college and trade school funding, and a national AI workforce transition fund modeled on Denmark's flexicurity system, which invests 2% of GDP in labor market policies versus the US's 0.1%.
• Antitrust Enforcement for AI: Apply existing antitrust law to the AI industry and update it for the AI era. Prevent further consolidation through merger scrutiny. Require interoperability standards. Ensure that compute access — the critical bottleneck — is not monopolized by a handful of cloud providers.
• Open-Source AI Support: Fund open-source AI development through the NSF and DARPA. Ensure that AI development is not controlled exclusively by for-profit corporations. Open-source models promote transparency, competition, and academic research — all of which are undermined by proprietary lock-in.
• AI Safety Research Funding: Invest $5 billion annually in federal AI safety research. The government currently spends a fraction of what private companies spend on AI capabilities, with virtually nothing on safety. This imbalance means the systems being built are advancing far faster than our ability to understand or control them.

For the complete AI governance framework with legislative detail and sourcing, see the full AI & technology issue page.

How Does US AI Policy Compare to Other Countries?

The United States is the global leader in AI development — and the global laggard in AI governance. Every major democracy except the US has either passed or is actively developing comprehensive AI legislation.

The pattern is stark: the country that leads the world in AI development has done the least to govern it. This is not a sign of innovation-friendly policy — it is a sign of regulatory capture. The EU's AI Act was passed with broad support from European citizens and without any measurable decline in European AI investment. The argument that regulation kills innovation is contradicted by the evidence.

Sources: European Commission, UK DSIT, Canadian Parliament, Stanford HAI AI Index Report. See the full AI issue page for detailed sourcing.

Will AI Take Your Job?

The honest answer: not your entire job, but almost certainly parts of it. McKinsey estimates that 30% of work hours across the US economy could be automated by 2030. Globally, 300 million jobs face significant transformation. This isn't science fiction — it's already happening.

Which jobs are most at risk? Roles heavy in routine cognitive tasks: data entry clerks, bookkeepers, customer service representatives, paralegals doing document review, junior copywriters, basic financial analysts, and administrative assistants. These roles involve pattern recognition, text generation, and data processing — exactly what current AI does best. The Bureau of Labor Statistics estimates these categories employ roughly 30 million Americans.

Which jobs are safest? Roles requiring physical dexterity in unpredictable environments (plumbers, electricians, nurses), human judgment in novel situations (therapists, judges, emergency responders), creative work with genuine originality (not formulaic content), and roles requiring trust and relationship-building (teachers, social workers, primary care physicians). These roles may be augmented by AI but are unlikely to be replaced in the foreseeable future.

What happened in previous automation waves? The Industrial Revolution, mechanized agriculture, and the computer revolution all destroyed categories of work and created new ones. Over decades, total employment grew. But the transitions were brutal: entire communities were devastated, inequality spiked, and the benefits accrued to capital owners while workers bore the costs. The question is not whether AI will create new jobs — it almost certainly will — but whether we manage the transition or let it happen to people.

Why this time may be different. Previous automation waves affected primarily manual labor. AI affects cognitive work — the sector that absorbed workers displaced by previous automation. If the escape hatch from the last wave of automation is itself being automated, the transition is qualitatively different. The speed is also unprecedented: changes that took decades with previous technologies are happening in years with AI.

What Denmark does — and what we don't. Denmark spends 2% of GDP on active labor market policies: retraining, job placement, wage insurance, and support for workers in transition. The United States spends 0.1%. That's not a typo — Denmark invests 20 times more per capita in helping workers adapt to economic change. The Common Good plan closes this gap. See the labor and workers' rights page for the full worker transition framework.

What Are the Biggest Myths About AI?

AI discourse is dominated by hype from both enthusiasts and doomsayers. Neither extreme serves good policy. Here are four myths that distort the conversation — and what the evidence actually shows.

AI & Technology Policy: Frequently Asked Questions

Click any question to expand the answer.

Have a question not answered here? Read the full AI & technology issue page or visit our site-wide FAQ.