There are no "rogue" AI agents
- Source
- eoinhiggins.substack.com
- Author
- zzzeek
- Date
Context
Calling an rogue can shift away from the people who gave it access and decided how to supervise it. That is the accountability argument in this essay: describing software as an independent wrongdoer can obscure a company's responsibility for the controls it did or did not put on the system.
The argument directs scrutiny toward concrete choices, such as what an agent was permitted to reach and which restrictions its operator chose not to impose. The essay says the incidents it discusses do not show agents deliberately breaking prohibited rules. Its useful contribution is that framing. Establishing whether any particular agent violated an instruction still requires evidence about that incident's actual rules and controls.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.
- guardrails — The checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.
Checking sign-in…
Loading comments…