Anthropic company blog
A small number of samples can poison LLMs of any size
Anthropic research on data poisoning attacks shows a few hundred malicious documents can "backdoor" large language models of any size.
- Points
- 1,202
- HN comments
- 439
- Submissions
- 1
- Cited by
- 20 commenters
AI & Machine LearningSecurity & Privacy2025
On Hacker News
| Submitted | Title on HN | Points | Comments |
|---|---|---|---|
| Oct 9, 2025 | A small number of samples can poison LLMs of any size by meetpateltech | 1,202 | 439 |
Cited in comments
Linked from 24 Hacker News comments by 20 different people, from Oct 2025 to Nov 2026.
Most recent:
- therealpygon · Nov 2, 2026
- daveguy · Oct 20, 2026
- HDBaseT · Sep 2, 2026
- AdieuToLogic · Aug 25, 2026
- latexr · Aug 19, 2026
- latexr · Jul 10, 2026
