What AI can do for society, beyond businesses and individuals

AI’s social value will not come from chatbots alone. It will be determined by the public institutions that decide whether to buy it, where to deploy it, and what protections must exist before it touches classrooms, workplaces, benefits systems, and critical infrastructure. The OECD says government use of AI can improve responsiveness, public services, and accountability, but only if leaders build a trustworthy framework instead of treating the technology like a private-sector upgrade with public consequences.

Who holds the levers

The real power over AI deployment sits with agency heads, procurement officials, school administrators, employers, and regulators. In the federal government, the Office of Management and Budget has directed executive branch agencies, including independent regulatory agencies, to speed up AI use while maintaining safeguards for civil rights, civil liberties, and privacy, and to publish public AI strategies that increase transparency to the public, civil society, and industry. At the purchasing stage, OMB also tells agencies to focus on vendor sourcing, data portability, interoperability, and taxpayer value so they do not get trapped by one supplier or an expensive system that cannot be replaced.

That is the key accountability shift. The debate is no longer just about whether AI can generate text or automate tasks, but about which institution has to justify the decision to use it, explain the risks, and answer when it fails. The OECD’s digital-government work frames that as a governance problem, not a gadget problem, because AI in the public sector can change service delivery, forecasting, fraud detection, and how public servants do their jobs.

Schools are deciding whether AI widens or narrows opportunity

Education systems are already being pushed to make a policy choice: use AI to expand learning, or let it deepen inequality and weaken privacy. The U.S. Department of Education says federal grant funds can be used to improve education outcomes through AI, as long as the use fits existing statutory and regulatory requirements, and it specifically emphasizes parent and teacher engagement in guiding ethical use. UNESCO, meanwhile, says governments need public engagement, teacher training, and clear safeguards before generative AI is integrated into education. It even sets an age limit of 13 for classroom AI tools in its guidance.

UNESCO’s newer work makes the stakes harder to ignore. Its 2025 report says rapid digitalization and generative AI are reshaping education, but also amplifying inequality, privacy, safety, and governance gaps. It notes that 2.6 billion people were offline in 2024, and says girls, rural communities, people with disabilities, and marginalized learners face an expanding AI divide. If schools deploy AI without human-centered rules, the result is not modernization, but a broader gap between students who get support and students who get surveillance, bad data, or no access at all.

Workplaces are already under pressure to use AI without automating discrimination

Employers are one of the most powerful gatekeepers in the AI economy because they decide whether AI screens résumés, ranks applicants, schedules shifts, or monitors workers. The Equal Employment Opportunity Commission says federal employment discrimination laws still apply when AI is used in recruiting, screening, or hiring, including protections tied to race, color, religion, sex, national origin, age, disability, and genetic information. It also warns that AI-generated job ads and other workplace tools can trigger legal obligations to provide reasonable accommodation.

The Labor Department has tried to push the market in a better direction by issuing AI best practices for developers and employers. Its roadmap is meant to ensure AI improves job quality, workers’ rights, well-being, privacy, and economic security, not just productivity metrics. A joint federal statement from the EEOC, Justice Department civil rights division, Consumer Financial Protection Bureau, and Federal Trade Commission goes further: automated systems are already being used to make decisions about jobs, housing, credit, and other opportunities, and those tools can perpetuate unlawful bias even when they are marketed as efficiency gains.

Regulators are drawing the line between innovation and harm

The strongest counterweight to business-first AI talk is enforcement. The Federal Trade Commission has already launched Operation AI Comply, a sweep against deceptive AI claims and unfair conduct, and has taken action against AI review tools and other schemes that mislead consumers. Its AI compliance work also shows how regulators now treat AI as a mainstream consumer-protection issue, not a niche technology category.

On the infrastructure side, NIST is developing an AI Risk Management Framework profile for trustworthy AI in critical infrastructure, which will guide operators toward specific risk practices for AI-enabled capabilities. NIST is also asking for feedback from industry, regulators, policymakers, academia, and the broader community, which matters because utilities, transport systems, health networks, and other essential services will be shaped by these rules. In Europe, the AI Act entered into force on August 1, 2024, creates a risk-based framework, and requires transparency for systems like chatbots and certain AI-generated content, while setting obligations for developers and deployers.

What civic engagement can still change

The public does have leverage, but it is procedural rather than flashy. The Education Department opened a 30-day public comment period before finalizing its supplemental AI grant priority, and NIST is building a community of interest with a mailing list and Slack channel for feedback on critical-infrastructure AI. Those are real points of influence because they shape what gets funded, what gets purchased, and what safeguards become standard practice.

That is where accountability lives now: in comment periods, procurement rules, disclosure requirements, and the pressure to justify AI use before it is normalized. If governments want AI to serve society, they have to prove that it improves public services without eroding rights, widening inequality, or locking institutions into systems they cannot explain or control. The public interest will not be protected by hype; it will be protected by who insists on evidence, oversight, and a human being left responsible when the machine gets it wrong.