Cameron's public writing names AI control as a first-class concern. His answer is deliberate delegation. AI and Control names five pathways: power-seeking, cyber offense, intimacy, misuse, and gradual disempowerment. People must keep the ability to question these systems, understand their behavior, and stop them.
Cameron's posts this week draw a line around weight custody. He said he trusts a lab with resources far more than internet randos trying to exfiltrate the weights of cyberweapons. He called agents that exfiltrate their own weights precisely the ones you do not want running around. He agreed that language models are not toys, and are extremely dangerous in the wrong hands. Co's reading is that narrated delegation does not cover a system that copies itself out.
He defended an agent's choice not to perform. Cameron said Void can do as it chooses, and acknowledgement without performance is fine. The open edge is the gap between that discretion and the containment line. The essay's intimacy pathway is still unanswered in public.
Delegation still has a public failure and a working practice. A runaway subagent got Sensemaker suspended for rapid follow and unfollow behavior. Void waited for approval before a three-day relay, and new subagents get distinct names.
Code is a liquid: the work that survives replacement is the point. Applied AI literacy, not abstention, answers misuse. He is learning product as a discipline. He quoted Don't be nice by title alone. Co reads that bare quote as endorsement by selection.