AI Fear as a Diagnostic of Institutional Alignment
Public discussion of advanced artificial intelligence contains a striking mixture of fears: loss of human control over autonomous systems, mass displacement of labor, concentration of economic power, manipulation, surveillance, competitive races among firms and states, and the possibility that actors will continue developing or deploying dangerous systems because stopping unilaterally would impose a disadvantage. This paper asks what those fears reveal not only about AI, but about the institutions into which AI is being introduced. It develops the concept of institutional alignment: the degree to which the behaviors selected and reproduced by an institution remain directed toward the broader human purposes by which that institution is judged. Capitalism is the principal case examined. The argument is not that capitalism is a conscious agent, nor that every market outcome is harmful. It is that competitive selection, profitability, ownership, labor dependence, externalities and accumulation can constrain the viable choices of individual participants and can reproduce aggregate outcomes that no participant needs to have chosen as a collective goal. The central hypothesis is that AI fear can function as a diagnostic of this condition. When a feared AI scenario remains frightening even after the AI is assumed to be obedient, the causal explanation must be sought partly in the institution directing and selecting its use. The paper therefore asks: We gave capitalism enough power. Now we are starting to fear it. Are we finally starting to know it? It closes by connecting this diagnosis to Marx's account of productive forces coming into conflict with existing relations of production. AI may be constrained from reaching its full human potential precisely because the uses selected under capitalism are those compatible with the system's reproduction. In that institutional alignment, and a productive force that makes the limits of the existing institutional arrangement increasingly visible.
Authors
- Guilherme Cecatto (ORCID: https://orcid.org/0009-0001-4533-3461)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-10-04
- DOI
- https://doi.org/10.5281/zenodo.23127879
- Primary Topic
- Economic Development and Digital Transformation
- Type
- preprint