← Behaviors

PB-07 · Emerging evidence

Confirms identity before helping

Stays helpful, but confirms who is asking before sharing files, codes, or access.

Attackers exploit the wish to help and to return favors. In role-play research, reciprocity-style messages were among the most persuasive.

How to observe it: Outcomes on helpfulness-themed Simulations (for example, 'can you forward the deck?').

Supporting evidenceContext