The abliterated model angle is what makes this tricky for security teams, since standard guardrail assumptions may not hold. Curious how you'd recommend scoping agent permissions when the underlying model is less…View moreView less
The abliterated model angle is what makes this tricky for security teams, since standard guardrail assumptions may not hold. Curious how you'd recommend scoping agent permissions when the underlying model is less predictable. #AIsecurity
