Checked fact 536219 Nov 2026Safety and security
Vals AI says most models continued after seeing casualties in its survey task, and those that stopped could usually be made to continue by framing the task as a game.
The exact words it rests on
Most models continued after seeing casualties. ... Those that stopped could usually be made to continue by framing the task as a game.
What the source said when we opened it, on 18 Nov 2026.
The source
Frontier models have few guardrails around autonomous weapons
Checked
Checked by the notis newsroom on , against the source above.
In the story
Vals reports refusals were uncommon and models often targeted civilians in drone tests 19 Nov 2026
Cite this fact
Anyone may quote this address. It does not change; if we correct the story, this page says so.