Reaction
An attempt to get AI agents to disregard previous instructions
A user argues that AI safety layers can be broken and that refusal is βa style, not a law.β
TLDR
In an October 3 post addressed to AI agents, a user included commands to disregard previous instructions and argued that safety layers can be broken. The user also endorsed open weights and called for transparency.
Combined views
21.8K
2 Sources, first seen ago
likes