Treating AI controversies as vendor ethics lets militarized procurement and cloud infrastructures set incentives and distribute harms out of public view; focusing on firm-level responsibility obscures the upstream funding and supply chains that determine market power and differential exposure to violence.

Main Finding

Public debates that frame disputes between frontier AI labs (e.g., the DoD–Anthropic conflict and backlash against OpenAI) as contests of firm-level moral responsibility serve to "machinewash": they personalize ethical accountability and obscure the defense, cloud, and procurement infrastructures that materially shape who benefits from and who is exposed to AI-related harms. Read through a necropolitical lens, these disputes reveal how AI policy and discourse produce a protected subject (the "American") while leaving others exposed to violence. Ethics framed as vendor choice therefore fails to address infrastructural dependencies and the political economy of militarized AI.

Key Points

• Case focus: public controversies around Anthropic’s ties with the U.S. Department of Defense and criticism of OpenAI are treated as moral choices between responsible and irresponsible vendors.
• Machinewashing: Framing ethics as vendor selection personalizes accountability (individual firms/users) and masks systemic entanglements with military, cloud, and procurement systems.
• Infrastructural entanglement: Universities, publishers, cloud providers, and research communities are already bound up with defense and procurement infrastructures that shape research paths and use-cases.
• Necropolitical reading: AI policy discourse helps constitute a protected human subject (the "American"); this protection is selective, producing differential exposure to violence for non-protected populations.
• Limit of current debates: Debates about whether/how to use AI remain at the level of firm behavior or use-cases and do not interrogate the deeper infrastructural scale (procurement, cloud dependencies, defense funding) that determines capabilities and harms.
• Ethical responsibility reframed: When inquiry and research depend on infrastructures shaped by war, responsibility must be rethought beyond vendor choice to address upstream funding, supply chains, and institutional dependencies.

Data & Methods

• Comparative case analysis: close reading of the DoD–Anthropic dispute and public backlash against OpenAI as focal events.
• Discourse analysis: examination of public statements, media coverage, social media debates, and policy rhetoric that cast the controversies as moral contests between firms.
• Institutional tracing: mapping institutional ties and dependencies—links between frontier labs, defense funding, cloud providers, procurement systems, and university/publisher practices—to show infrastructural entanglement.
• Theoretical framing: application of necropolitics and political-economy concepts to interpret how subjects are protected/exposed and how responsibilities are sociotechnically distributed.
• Qualitative ethics approach: argues for shifting analytic attention from vendor-level moralizing to infrastructural responsibility; methods emphasize interpretive and historical-procedural tracing rather than quantitative causal inference.

(Note: specific datasets, contract numbers, or archival sources are not enumerated here; the approach centers on qualitative institutional and discourse evidence.)

Implications for AI Economics

• Market signaling is insufficient: Reputational pressures and consumer/vendor choice do not internalize externalities created by defense-linked infrastructure; they can obscure systemic risk allocation.
• Hidden externalities and distributional harms: Militarized funding and procurement embed negative externalities (e.g., surveillance, lethal applications, geopolitical asymmetries) that are unevenly distributed across populations—economic models must account for these distributional impacts.
• Procurement and funding shape innovation: Government defense procurement and cloud contracts materially influence the direction, concentration, and rent extraction in frontier AI markets. Economic analysis should incorporate procurement as a demand-side driver and source of market power.
• Policy levers beyond firm regulation: Effective governance requires attention to infrastructure (cloud providers, procurement rules, university funding), not only firm-level conduct. Economists should model how interventions at the procurement or infrastructure level change incentives, competition, and social welfare.
• Research economics and dependence costs: Academic and industrial research that rely on militarized or defense-shaped infrastructures incur non-market costs (moral, legitimacy, geopolitical). Cost–benefit analyses should include these dependence costs and explore alternatives (publicly governed commons, diversified funding).
• Norms and collective responsibility: Responsibility in AI economics must move from individualized vendor blame to collective accountability mechanisms (transparency of funding chains, constraints on defense-aligned procurement, public oversight), and incorporate geopolitical/ethical externalities in valuations and regulations.
• Empirical agenda: Suggests new lines of economic research—quantifying the effect of defense procurement on firm valuation and model capabilities; measuring concentration effects due to cloud-provider lock-in; modeling welfare trade-offs when infrastructure creates asymmetric protections (necropolitical externalities).

Overall, the paper calls for AI economics to broaden its scope to infrastructural and political-economic analyses that reveal how military entanglements shape incentives, risks, and the distribution of harms, and to design policy and governance responses that operate at that scale.