Anthropic's answer on Claude and the Borisov-Gabber-Vasiu paper sidesteps the key question

giffmana · x · 2026-09-10

Following questions about whether Claude accessed the (accidentally) publicly available Borisov-Gabber-Vasiu paper on the Jacobian conjecture, an official response was issued — but as Suchenzang notes, it doesn't answer whether the paper made it into Claude's training data or figured in its reasoning, raising the question of whether such provenance is even detectable. giffmana says he'd recently thought the exact same thing.

Related event: Anthropic responds to Jacobi conjecture data leak doubts(2 posts)→

Original post →

More from Models

Models channel →