Amazon says human-in-the-loop AI oversight is failing because humans stop paying attention — 2026-06-21

## Short Segments Waymo's expansion into New York City hits a roadblock, and it's not about technology. The taxi lobby and political opposition have halted Waymo's entry into the Big Apple, despite its success in other cities. Coming up, we'll explore how Amazon's critique of human-in-the-loop AI oversight could reshape AI governance. Waymo's ambitions to operate in New York City have been thwarted by political and labor opposition, not technical challenges. The company, which provides over 500,000 rides weekly across ten U.S. cities, faces resistance from local politicians and the taxi lobby. New York Governor Kathy Hochul recently withdrew a proposal that would have allowed robotaxi services, citing a lack of support. This political pushback highlights the influence of traditional taxi services and labor unions in shaping urban transportation policy. For Waymo, this means its expansion plans in New York are on hold, underscoring the complex interplay between innovation and regulation in the autonomous vehicle sector. ## Feature Story Amazon challenges the effectiveness of human-in-the-loop AI oversight, arguing that human attention is unreliable. Eric Brandwine, Amazon's VP of Security, claims that relying on humans to monitor AI systems is flawed due to human inconsistency and the normalization of deviance. This perspective is gaining traction among tech giants like Google, Microsoft, and IBM, who are reconsidering the role of human oversight in AI governance. Brandwine's critique centers on the idea that humans, much like AI, can become desensitized to deviations from expected behavior, leading to oversight failures. This phenomenon, known as normalization of deviance, suggests that humans may gradually accept unsafe practices as normal. The implications are significant for industries deploying AI at scale, as it questions the reliability of human oversight in maintaining safety and ethical standards. As companies like Amazon and others in the tech industry reassess their AI governance strategies, the focus may shift towards more automated and robust systems that minimize human intervention. This could lead to a reevaluation of current AI deployment practices and potentially accelerate the development of more autonomous AI systems. The debate over human-in-the-loop oversight is likely to continue, with significant implications for how AI is integrated into critical operations across various sectors.