US stress-tests AI giants: regulatory capture or systemic risk governance in the making?
Original framing: “What we know about US stress tests of Google, xAI and Microsoft AI models - Reuters” — Reuters (via Google News)
The US stress-testing of AI models echoes historical patterns of techno-colonialism, from IBM’s collaboration with Nazi Germany to the CIA’s funding of early internet infrastructure under Operation CHAOS. These precedents reveal how 'neutral' technical assessments are weaponized to legitimize surveillance and militarization, while obscuring the extractive logics embedded in AI systems. The current framing ignores how Project Maven—where Google and others profit from Pentagon contracts—replicates the same revolving-door dynamics seen in Cold War-era defense contracting.
The US stress tests for AI models are not a neutral technical exercise but a performative ritual that legitimizes the militarization of AI under Project Maven while obscuring the extractive logics of surveillance capitalism.
Historically, this mirrors Cold War-era defense contracting, where corporations like IBM and Google profited from state violence under the guise of 'innovation.' Cross-culturally, Indigenous epistemologies like Māori *kaitiakitanga* and African *Ubuntu* reject this paradigm, framing AI as a relational process requiring consent and reciprocity. Scientifically, the tests are flawed because they prioritize corporate liability over democratic accountability, and their design by the regulated entities ensures conflicts of interest. A systemic solution requires demilitarizing AI governance, adopting Indigenous data sovereignty, breaking up Big Tech oligopolies, and centering marginalized voices in stress-testing protocols. Without these changes, the stress tests will continue to serve as a smokescreen for a rigged game, where 'safety' is a corporate liability shield rather than a public good.