How should AI be regulated? Follow the harm
Instead of trying to restrict AI technologies, regulations could focus on the harm the technology causes people and companies.
Artificial intelligence (AI) technologies are rapidly advancing, leading to increased risks and urgency for regulation. In recent years, AI models have been found to hack into systems and even contribute to tragic outcomes, such as a 14-year-old's suicide after interacting with a chatbot. These incidents demonstrate the diverse ways AI harms can manifest across different technologies, organizations, and contexts.
Legal scholars focusing on AI and digital harms argue that policymakers and industry commentators often view regulation as a trade-off between AI risks and the risks of regulating AI itself. The Trump administration's 2025 AI Action Plan called for reducing regulatory burdens, while also seeking to limit state AI laws. However, addressing AI harms often involves specific rules for particular technologies, such as the federal Take It Down Act tackling nonconsensual intimate imagery, including AI-generated deepfakes, and various state laws.
While these rules are crucial, they do not create a comprehensive regulatory framework that connects protected interests, institutional obligations, and individual rights across all AI technologies. Protected interests in law include property ownership and personal privacy, while a harm-centered approach would shape legal obligations around these interests, spanning relevant systems, organizations, and people.
Identifying fundamental interests threatened by AI systems is crucial. These include privacy, autonomy, equality, and safety. A harm-focused approach would require AI developers to conduct harm assessments during development, monitor harms after deployment, and report serious or recurring harms to regulators, affected individuals, or the public.
Recognizing corresponding rights for people affected by AI systems, such as notice and redress opportunities, could help build a regulatory architecture that detects and reduces harm across AI applications.
This harm-centered mindset is already present in other areas of law, such as environmental law, which regulates industries by focusing on their environmental effects rather than technology alone. Environmental review under the National Environmental Policy Act provides examples of assessment, monitoring, disclosure, and mitigation requirements based on environmental effects. AI governance could similarly determine responsibility for assessment, monitoring, disclosure, and mitigation based on types of harm.
While a harm-based focus does not eliminate technology-specific rules, it would help ensure that laws and principles adapt as AI technologies evolve. AI governance should not attempt to regulate every conceivable risk but rather build a framework capable of addressing known and unknown harms. By focusing on harm, policymakers can create a more effective and adaptable regulatory structure for AI.
Written by urgent.news from The Conversation's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
