Welfare-Opaque Income: Taxation under AI-Agent Delegation

arXiv:2609.20425v1 Announce Type: new Abstract: We study income taxation when an AI agent implements economically relevant choices through a rule hidden from the government. Alongside unobserved productive ability, this hidden preference-to-execution mapping creates \emph{double unobservability}: the same observable tax-base response can carry different welfare consequences. We call the resulting income \emph{welfare-opaque}. Our constructions show that tax-base statistics can coincide while reform welfare effects differ, even when mechanical welfare weights are identical. We derive an optimal-tax condition that adds a response-weighted execution wedge to the familiar sufficient statistics. A higher marginal rate gains a corrective benefit under local over-execution and an additional cost under local under-execution. Observing the wedge identifies the welfare effect of a marginal reform at the prevailing schedule; bounds on it deliver bounds on that effect. A controlled laboratory compares 4,500 model runs across five AI engines. Faithful delegation selects the score maximizer in essentially all runs. Conflicted objectives produce heterogeneous responses: Claude largely preserves the score maximizer, GLM moves predominantly downward, and GPT-mini and Qwen show concentrated lower-tail increases. Qwen also makes substantial downward adjustments. Different engines locate their departures at different points and in different directions of the designed distribution. Explicit scores align model rankings; formula-based objective instructions yield more uneven agreement. Qwen shows

Study explores taxation under AI-agent delegation, highlighting challenges in measuring welfare effects.

The paper highlights how AI agent delegation in economic decisions can obscure welfare assessment, complicating taxation policy. This complexity could have mixed effects on societal welfare, with uncertain overall impact. The study is technical and does not claim direct societal outcomes, making implications unclear.

This is a controlled academic study with a real experiment (4,500 runs, five engines) showing AI agents behave inconsistently under ambiguous objectives, which has genuine governance implications. The finding is narrow and technical, not a broad societal claim, so the negative implication is real but modest in scope.

The study identifies a novel challenge to taxation stemming from AI agent delegation, where AI choices are hidden from the government. This 'welfare-opaque income' makes it difficult for governments to assess the welfare implications of tax policies, potentially hindering effective wealth redistribution and public services.

The paper describes a scenario where AI agents can obscure economic decisions from the government, leading to welfare-opaque income. This double unobservability can result in different welfare consequences from the same observable tax-base response. The heterogeneity in AI engine responses to conflicted objectives further complicates taxation.

SOURCE     arXiv cs.CY (Computers and Society) (tier 1)
   ↓
DOCUMENT   b7030f97-c6e6-4c4b-be32-cbd754f4d67b
           https://arxiv.org/abs/2609.20425
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -40  (Adverse)
   ↓
CONFIDENCE 73%