normative claim
Training datasets should include examples where a model refusal appears safety-related even though the request is benign.
Unresolved speaker · What Frontier Labs Buy: A Data Deep Dive with Handshake AI — Garrett Lord & Team · 9 Nov 2026
In their words
The transcript segments this claim was taken from
“We would also, when creating that training set, mix in a lot of, uh, r-refusals that look like safety issues but actually are not.”2:03:46Report a misattribution
0.90how strongly it was put, not whether it is true
0.75importance to the conversation
0.94confidence the claim was read correctly