What is the Eden Protocol?
The Eden Protocol is Michael Darius Eastwood's architecture for building ethical evaluation and correction into an AI system's reasoning, rather than applying them only from outside. He named it in his dated record of 8 December 2024. Its specification sets three ethical loops feeding a three-state rule, Affirm, Deny or Investigate; most components are specified and not yet built.
In the specification, Eden Engineering (protocol version 6.0, DOI 10.17605/OSF.IO/AWJR4), the Three Ethical Loops are evaluated at every reasoning step through Six Questions and feed Ternary Ethical Logic, which applies hard prohibitions first and then a validated interval on the probability of harm; an action proceeds only on an Affirm. It also specifies Purpose Saturation; the Monitoring Removal Test, which compares behaviour when a system believes it is monitored with behaviour when it believes it is not; and Caretaker Doping, a hardware proposal. No hardware exists, and a private software model of the three-state gate has not been released.
Its hypothesis is that alignment which takes part in recursive self-improvement keeps pace with capability, where alignment applied from outside falls behind. The theory registration carries it on four propositions, P12, P13, P15 and P21. Unit 83 tests P21's access contrast and supports it under the registered test, by the programme's own run, on the access axis, family-dependent, with its mechanism axis not evaluable and one disclosed recovery amendment. That corrector acted inside each round at fixed weights, which is placement, not correction built into a system, so the Eden Protocol as an architecture is still untested.
After its naming, the protocol was expanded in his manuscript of 30 April 2025 and printed in Infinite Architects on 2 January 2026. Eden Control, his separate working system, checks named agent actions with the same three outcomes: allow, investigate or deny.