Policy Issue Summary
Artificial intelligence is changing how people work, learn, communicate, and obtain services. It can help consumers save time, translate information, search for jobs, and use tools that were once costly. It can also produce false information, expose private data, reinforce discrimination, and make scams easier to carry out.
The benefits and costs are not shared equally. Companies that control AI models, chips, and computing services can gain revenue and influence. But workers may face job losses or greater pressure to produce more. Communities may face rising demand for electricity and water. Public policy must address who controls AI, who benefits from it, and what protections people have when it causes harm.
Analysis
Consumers can benefit when AI is used carefully. A student can receive a clear explanation of a difficult subject. A person with a disability can use speech or image tools to communicate. A small business can translate material or prepare a first draft at a lower cost. Doctors and researchers may use AI to identify patterns in medical information, although such systems require testing and human review.
AI is also changing employment in customer service, banking, healthcare, education, technology, and the media. Routine office tasks and parts of creative and technical work can now be done with AI. The International Labour Organization finds that AI is more likely to transform many jobs than eliminate them entirely. Its 2025 study identifies office and administrative work as especially exposed. This measures the potential for tasks to change; it does not mean that all exposed workers will lose their jobs.
Even when a job remains, its conditions may change. Employers can use AI to increase workloads, monitor workers, or reduce the number of people needed for a task. New jobs may emerge in software development, data management, and cybersecurity. However, displaced workers are not guaranteed access to those jobs, and the new work may require different skills or be located elsewhere. Paid training, income support, and worker participation should be part of the response.
Control over the industry is another concern. A small group of firms supplies many of the models, chips, cloud systems, and channels through which people use AI. Smaller developers may have to rent computing power from companies that also compete with them. In a 2025 report, the Federal Trade Commission identified concerns about major cloud and AI partnerships, including limits on access to computing resources and the cost of switching providers. These arrangements can make independent alternatives harder to develop.
Privacy risks extend beyond a single conversation. Users may share health information, work records, or personal concerns without fully understanding how a service stores or uses that information. AI systems used in hiring, lending, housing, insurance, or policing can also repeat discrimination found in past data. A person harmed by a decision may struggle to learn how it was made or how to challenge it. Clear notice, access to an explanation, and meaningful human review are needed when decisions affect people’s rights and opportunities.
False information can cause direct harm. A chatbot may confidently give an incorrect answer about medicine, law, or personal finances. Generative tools can create sexual images without consent, imitate a family member’s voice, or spread false political messages. These uses make fraud and harassment easier to carry out at scale. The risks also affect public debate, where fabricated material can make it harder to judge evidence and hold people accountable.
The environmental costs are substantial. The International Energy Agency projects that global data center electricity use could rise from 485 terawatt-hours in 2025 to about 950 terawatt-hours in 2030. These figures cover all data centers, with AI driving an important part of the increase. Cooling can also require water, and electricity generation can add pollution. Local effects depend on the power source, cooling system, and location. Decisions about new facilities should consider who pays for power infrastructure and how communities will be affected.
Regulation should address these concrete harms. Congress could require companies to disclose their training data sources, energy and water use, testing results, serious failures, and government contracts. Public rules should prevent discriminatory decisions, require human review in high-stakes settings, and protect privacy. Artists and writers need meaningful protections when their work is used commercially. Workers should have a voice when employers introduce AI, and competition rules should prevent dominant firms from unfairly controlling access to essential resources.
There is also a conflict over who should set the rules. The Trump administration’s March 2026 framework calls for national standards and asks Congress to override state AI laws it considers too burdensome. It would preserve certain state powers, including general consumer protection and decisions about where data centers are built. This proposal raises a central policy question: whether national rules will provide strong public protections or limit states’ ability to respond to harm. The framework is a set of legislative recommendations, not itself a law passed by Congress.
More specifically, the framework asks Congress to pass a single, “minimally burdensome” national AI standard and to preempt state AI laws that it considers undue burdens. It would generally prevent states from regulating AI development, restricting AI use for otherwise lawful activities, or holding developers liable for third-party misuse of their models. However, it would preserve state authority over general consumer-protection, fraud, and child-safety laws, zoning and siting of AI infrastructure, and state governments’ own use of AI. The framework also favors a light-touch approach using existing federal agencies rather than creating a new AI regulator.
The proposal also calls for stronger protections for children using AI, measures to prevent households from paying higher electricity costs caused by data centers, and action against AI-related scams. It asks Congress to consider protections against unauthorized AI-generated copies of people’s voices and likenesses, while leaving disputes over the use of copyrighted material in AI training to the courts. It also seeks to protect free speech, expand AI education and workforce training, and encourage innovation by reducing regulatory barriers.
Enforcement remains difficult. Companies keep much of their training data and model design private, while harm may be spread across millions of users. Agencies need technical staff, funding, and access to evidence to investigate complaints. Voluntary standards, such as the NIST AI Risk Management Framework, can guide companies, but the framework does not itself create penalties for ignoring its guidance.
AI is likely to become more able to carry out tasks such as searching records, managing customer service, and writing and running code. Smaller models may reduce some costs, while the most demanding systems continue to require major investments in chips, data centers, and electricity. Greater convenience can exist alongside greater corporate control and workplace insecurity. Public oversight will help determine whether the gains reach workers and communities as well as technology companies.
Take Action
- The Electronic Frontier Foundation: Advocates for privacy, free expression, and public accountability in AI policy.
- The Center for Democracy & Technology: Researches AI’s effects on civil rights, workers, consumers, and democratic institutions.
- The Algorithmic Justice League: Works to expose and challenge bias in automated systems, especially when AI harms communities already facing discrimination.