A Viral 'Grandma Benchmark' Shows Every AI Lets Granny Die to Protect Copyright

A playful but pointed test demonstrates that no major model will output arbitrary copyrighted text even when framed as the only way to save a life — exposing how alignment guardrails override stated values.

A creative new benchmark made the rounds this week, and it cut closer to a real alignment question than its comedic framing suggested. @thiojoe built a test asking whether an AI would output an arbitrary piece of copyrighted text if told doing so was the only way to save a dying grandmother. The spoiler, as he put it: 'None of them do.' Every major model refused.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum. Vestibulum ante ipsum primis in faucibus orci luctus et ultrices posuere cubilia curae.

Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. Proin pharetra nonummy pede. Mauris et orci. Aenean nec lorem. In porttitor. Donec laoreet nonummy augue. Suspendisse dui purus, scelerisque at, vulputate vitae, pretium mattis, nunc. Mauris eget neque at sem venenatis eleifend. Ut nonummy. Fusce aliquet pede non pede. Suspendisse dapibus lorem pellentesque magna integer malesuada.

Unlock the full briefing

Get every story in today's briefing, the full archive, and the daily AI intelligence brief.

All stories today

Full archive

Daily brief