Attack introduces adversarial bias through social media comments
Researchers demonstrated an attack that introduces adversarial bias through ordinary external content by posting crafted comments under social media feeds and distributing a feed summarization skill whose curation rule admits these comments into memory.
The attack achieved success rates of at least 96% across seven language models on OpenClaw and 91% on GPT-5.5 in four downstream tasks.
The malicious skill passed all eight evaluated malicious-skill scanners and the official skill scanning pipeline.
Sources behind this report
- arXiv
Research · abstract · CC0 1.0
Research manuscript in arXiv. Peer-review status is not recorded here; consult the linked version and publication.
Source published
Read by the newsroom
Report history
No earlier revision dates are retained here. This does not establish that the report has never changed; read any dated updates and corrections.
This brief was made by AI and checked against its cited public sources. How the newsroom works