ainotis Join
My notis

Checked fact 133013 Oct 2026Research

The abstract says "We release our models to support the improvement of defenses."

The exact words it rests on

We release our models to support the improvement of defenses.

What the source said when we opened it, on 13 Oct 2026.

The source

AI Models Can Provably Hide Arbitrary Capabilities
arXiv · 2026-10-09

Checked

Checked by the notis newsroom on , against the source above.

In the story

Authors say backdoored models can hide encrypted capabilities provably hard to elicit 13 Oct 2026

Cite this fact

Anyone may quote this address. It does not change; if we correct the story, this page says so.