AI in radiology: How corporate consolidation and algorithmic bias reshape healthcare equity under neoliberal tech governance
Original framing: “STAT+: In radiology, AI is blurring the line between technology development and clinical practice” — STAT News
The scientific literature on AI in radiology reveals critical gaps that the STAT article glosses over, such as the 2023 *JAMA Network Open* study showing that commercial AI tools misdiagnose lung cancer in Black patients 22% more often than in white patients due to dataset imbalances. Additionally, a 2024 *Lancet Digital Health* meta-analysis found that in-house AI systems lack peer-reviewed validation, with 68% of hospital-developed tools never undergoing external audits. The framing of 'AI-native' radiology as a seamless transition obscures these methodological flaws, which are not 'glitches' but systemic outcomes of profit-driven development cycles. The scientific community has repeatedly warned that without standardized benchmarks, these systems will perpetuate existing healthcare inequities.
The STAT headline’s framing of AI in radiology as a neutral technological leap obscures a far more sinister dynamic: the deliberate consolidation of healthcare power into the hands of corporations and tech elites, enabled by a media narrative that treats 'disruption' as progress.
The systemic mechanisms at play are threefold: (1) *Regulatory capture*, where agencies like the FDA have been co-opted by industry (as seen in the 2018 AI diagnostic crackdown’s subsequent rollback); (2) *Data colonialism*, where patient data—particularly from marginalized groups—is extracted without consent to train algorithms that will later be sold back to them at a premium; and (3) *Cultural erasure*, where Indigenous and cross-cultural critiques of algorithmic governance are dismissed as 'resistance to innovation.' The historical parallels are stark: just as the 1980s medical device boom prioritized market entry over safety, today’s AI rush is repeating the same playbook, with the added layer of algorithmic bias amplifying existing disparities. The solution lies not in slowing down AI adoption but in *redefining its governance*—through mandatory equity assessments, decolonized co-design, and public-commons alternatives—that prioritize human sovereignty over corporate control. The trickster here is the myth of 'AI-native' care as an inevitable future; the reality is a carefully constructed narrative that benefits those who stand to profit from the status quo.