AI rules often target particular technologies, but legal scholar Sylvia Lu argues they should also follow the harms systems cause. Her proposal would tie duties for developers and deployers to risks such as privacy violations, discrimination and threats to safety or autonomy.
Watch Desk analysis
What happened
Lu, an assistant professor of law at American University, argues for a framework that starts with protected interests and assigns responsibilities across the people and organisations involved in AI. Rather than relying only on rules for particular tools, she proposes that developers assess foreseeable harms during development and explain how they tested for and mitigated them.
Deployers, including companies and government agencies, could have duties to monitor harms after launch. Serious or recurring harms might trigger reporting to regulators, affected people or the public, while people affected by AI could have rights to notice and, where appropriate, redress. Asia Times republished Lu’s essay from The Conversation.
Why it matters
The proposal connects safeguards to what an AI system does to people, not just what kind of system it is. That could make the rules more adaptable as technology changes, while putting continuing responsibilities on organisations after deployment, not only during development.
Lu does not suggest abandoning technology-specific rules. Requirements for deepfakes or high-stakes decisions could still be needed; a harm-based framework would sit alongside them. The hard question is how to define a harm clearly enough to make duties enforceable without treating every imaginable risk as grounds for regulation.
Our read
This is a useful organising principle, not a ready-made rulebook. Its strength is the link between assessment, monitoring, disclosure and redress; its test will be whether lawmakers can turn that chain into obligations people can actually use. “Follow the harm” is a fine slogan. The unglamorous definitions and enforcement are where it earns its keep.
What to watch
- Whether lawmakers propose specific duties for assessing and monitoring AI-related harms.
- How serious or recurring harms would trigger reporting, and to whom.
- Whether affected people would gain practical routes to notice and redress.
Discussion spark: Should AI regulation begin with the harm people suffer, or do developers and deployers need technology-specific rules first?
Sources and evidence
- How to regulate AI? Follow the harm – Asia Times (28 October 2026, 06:08 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.