← Latest AI News

November 3, 2026 · Daily brief

Proton canceled AI job reviews after staff leaked the push

Sovereignty angle
When your boss measures how much AI you use instead of what you shipped, you are being graded on hours in a tool, not results. That is the wrong number to count.

Proton CEO canceled company-wide AI usage interviews one day after reports showed he dismissed security testing as loser thinking.

Proton CEO Andy Yen canceled planned company-wide re-interviews about AI usage on October 30, one day after a report showed he dismissed a staff request for security testing as loser thinking. The company had planned to interview nearly all employees between November 4 and 11 about how much they use AI. Proton CTO Bart Butler said in a meeting that 90 percent of the code the company wrote in October was generated by AI, largely using Anthropics Claude. Employees told reporters they believed people who used AI too little could be fired. After the backlash, Yen canceled the interviews but said the same AI usage metrics would become part of regular performance reviews.

What you can do with it now

If your manager starts asking how much time you spend in ChatGPT or Claude, shift the conversation to results. Keep a weekly log of what you shipped with AI help: three proposals sent, 12 follow-up emails written, one hiring post drafted. When the question comes up, answer with output, not hours. If pressure builds to use AI on tasks where it makes the work worse, document one example where you tried the AI version and it failed a client check. That record protects you if usage becomes a performance metric.

What to keep in your own hands

Never paste client names, invoice details, or confidential project data into an AI tool just to hit a usage target. The Proton story shows what happens when speed becomes more important than checking the work. If your workplace installs software to track AI usage, ask in writing what data it collects and whether it logs the content you send to AI providers. That question alone often slows the rollout. Keep one task you do manually as a benchmark. If AI truly makes you faster, the manual version will feel slow by comparison. If it does not, you have proof the tool is not helping.