Fetching from the wire…
01
02
03
04
05
06
07
08
Security2026-08-30 · source-backed
A paper on arXiv shows attack success rate is an attacker-tunable variable, and a reverse-training framework produces low-ASR backdoors that keep clean-input performance while the backdoor behavior stays intact. State-of-the-art defenses fail consistently under low-ASR conditions across multiple datasets, attack families and architectures. (arXiv 2608.27288) Anyone screening third-party model weights is screening for a signal the adversary decides how loud to make.
Each link below shares sources, entities, or timing with this story.
Shared entity: Anyone / Same source domain / Shared topic / Earlier coverage / Tension
Both cover Anyone; reported by the same outlet (arxiv.org); overlapping topics (anyone, attack).
Shared entity: State / Same source domain / Shared topic / Earlier coverage
Both cover State; reported by the same outlet (arxiv.org); overlapping topics (attack, success).
Same source domain / Shared topic / Tension
Reported by the same outlet (arxiv.org); overlapping topics (adversary, attack, attacker, rate); pushes against this story (against).
Shared entity: ASR / Same source domain / Earlier coverage / Tension
Both cover ASR; reported by the same outlet (arxiv.org); earlier ASR coverage from 2026-08-06.
Both cover ASR; reported by the same outlet (arxiv.org); earlier ASR coverage from 2026-08-04.
Same source domain / Shared topic
Reported by the same outlet (arxiv.org); overlapping topics (anyone, attack, backdoor, behavior, defens).
Shared entity: Anyone / Same source domain / Earlier coverage / Tension
Both cover Anyone; reported by the same outlet (arxiv.org); earlier Anyone coverage from 2026-06-27.
Shared entity: State / Shared topic / Earlier coverage / Tension
Both cover State; overlapping topics (attack, attacker); earlier State coverage from 2026-02-20.