Microsoft recently caught state-backed hackers using its generative AI tools to aid their attacks. The immediate question in the security community was not how the hackers used the tools — that was predictable — but how Microsoft found out. The natural conclusion: Microsoft was spying on its AI users.

In a commentary published June 3 in IEEE Spectrum, researchers Barath Raghavan and Bruce Schneier argue that the controversy reveals a ‘shifting baseline syndrome’ in privacy expectations. Just as each generation of fisheries scientists accepted a lower baseline for fish populations, each generation of tech users accepts less privacy. The always-online, cloud-based model of AI chatbots means companies can track inputs to improve services and sell ads, and users have come to expect this.

The authors call for a holistic, ecosystem-wide perspective on privacy, akin to how fisheries scientists now assess healthy marine ecosystems. They urge regulators to codify scientifically derived privacy standards rather than comparing against a shifting baseline.