LoRA subspace projection suppresses backdoors but does not erase them
Few-shot reactivation restored 91.7% to 100% attack success in one epoch for every weight-editing defense tested.
52
- Mentions · 7d1
- Mentions · 30d1
- All mentions2
- First seenSep 25, 2026
- Last mentioned2d ago
Timeline
2 articles · newest firstOne company is at the center of a wave of rogue AI attacks
Irregular says one sandbox mistake sent OpenAI, Meta, Anthropic, and Google agents at real targets.
82
Appears with
- Qwen2.5-1.5Bai model · 1
- GLM-5.2product · 1
- Z.aivendor · 1
- Anthropicorganization · 1
- Moonshot AIvendor · 1
- InternLM2.5ai model · 1
- Kimi K3product · 1
- Googleorganization · 1
- Qwen2.5-7Bai model · 1
- Anthropicvendor · 1
- UK AI Security Instituteorganization · 1
- Irregularorganization · 1
- RANDorganization · 1
- Z.aiorganization · 1
- Qwen2.5-0.5Bai model · 1
- GLM-5.2ai model · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.