Where intention goes, responsibility flows: Grounding AI accountability in the machine learning paradigm.

Machine learning systems are built from human-specified objective functions plus traceable human decisions across the pipeline, making AI output a composite of layered intents rather than one speaker.

• Why it matters: US lawsuits are starting to hold AI companies accountable, but existing legal routes and immunities are uneven, leaving courts without a clear, doctrine-neutral way to assign responsibility.
• What they did: The Perspective grounds accountability in ML engineering by tracing intentional design choices across training, deployment context, user interaction, and postdeployment oversight, then applying three doctrine-neutral questions on attribution, causation, and apportionment.
• The result: It identifies four remaining gaps—shared responsibility, proving fault in opaque models, postdeployment harms, and fast remedies—and argues targeted regulatory reforms can address them without new legal categories or AI personhood.

AI summary of the abstract · may contain errors · not medical advice · read the paper.