Role Grounding Attack against Agent Memory via Social Interaction

# Role Grounding Attack against Agent Memory via Social Interaction

https://gcdradar.vercel.app/incidents/role-grounding-attack-against-agent-memory-via-social-intera · AIR-2026-1695 · closed 2026-11-13

- Lane: Research
- Severity: 1 (Research / proof of concept): Research paper demonstrating an agent memory manipulation attack.
- Fix status: Fix status unknown
- First reported: 2026-10-30
- AI role: Attack on AI
- Products: LLM-based Agents

## Summary

Researchers demonstrate Role Grounding Attack, a vulnerability in LLM agent memory systems where malicious social interactions cause an agent to misattribute attacker preferences to the user. The technique exploits role-to-principal attribution flaws in multi-agent or social environments.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-10-30: Is it my preference? Role Grounding Attack against Agent's Memory via Social Interaction (https://arxiv.org/abs/2610.38209)

## Sources

- 2026-10-30 arXiv: Is it my preference? Role Grounding Attack against Agent's Memory via Social Interaction https://arxiv.org/abs/2610.38209

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.