Li Dajun, Deng Tianqi
Journalism Evolution.
2026, 15(1):
35-43.
In the intelligent communication environment where news organizations and digital platforms operate jointly, generative artificial intelligence has significantly improved content production and interaction efficiency, yet it has also given rise to the hidden and accumulating risk of “AI sycophancy.” Based on a yearlong investigation of mainstream media organizations, internet companies, content governance platforms, and news practitioners, this study employs multiple case studies and grounded theory, and constructs a fourstage analytical model—“driversmanifestationsrisksgovernance”—within the framework of risk society theory. The findings indicate that the drivers of AI sycophancy stem from the combined effects of model training biases, user interaction cues, and platform governance logics; its manifestations exhibit stable structures across content softening, stance following, and value alignment; its risks accumulate and spill over across the three dimensions of news practice, platform information, and governance systems. Correspondingly, governance pathways have gradually formed a systematized framework comprising technical governance, platform governance, and organizational governance. The analysis provides a new theoretical lens and practical reference for understanding and regulating institutionalized soft risks in the context of generative artificial intelligence.