Greg Twemlow argues for the urgent need for a constitutional framework to govern the rise of autonomous AI agents, termed "Agentic AI." It posits that this new form of power, unlike atomic or digital power, rewrites reality by delegating decision-making to synthetic entities, which can recursively improve and operate faster than human governance. The author proposes MAARR (Memorandum of Agentic AI Rights and Responsibilities) as a constitution for the age of agentic intelligence, outlining principles like agent auditability, rights allocation, and sovereignty safeguards. Enforcement mechanisms are discussed, including an Agentic Firewall that validates compliance at the execution layer and the importance of GPU vendors integrating MAARR compliance. The article concludes by highlighting the potential for a future conflict between MAARR-enabled democracies and "Shadow MAARR" architectures, underscoring that MAARR is not optional and must be embedded into the core of synthetic cognition. Read the article.

Ver todos los episodios