AI and health data: trust must be built in, not promised

Illinois Governor JB Pritzker says government must become more aggressive in regulating AI and big technology to protect people from concentrated power and intrusions into their privacy, and that AI carries potentially existential risks and large economic effects in the hands of too few companies (The Verge).

In Switzerland the question has a concrete shape. On 15 September 2026 the National Council approved a new electronic health record by 134 to 53: every resident would get one automatically and free of charge, with the right to object or have data deleted (inside-it). Critics warn that the bill does not ensure that only the people concerned hold the keys, and ask whether central health data would invite cyberattacks.

What one side asks of regulators, the other asks of design: trust must be built in, not promised. AI will help in health care, with documentation, research and relief for staff. But the criticism is also the checklist for any hospital or practice that connects an AI tool to patient data: keys held by patients, logs of every access, a clear opt-out.

Many health systems are short of staff, and AI documentation tools need staff who can use them safely. Training people from other sectors for health administration roles is one of the clearest paths from job loss to a stable new job.

Practical step: before connecting an AI tool to patient or client data, ask who holds the keys, who can see the logs and where the data is processed.