Microsoft AI releases draft Humanist AI Code of Conduct

A September 14, 2026 post says Microsoft AI published 10 draft principles for its own models but had not yet used them to train current models. The summary says evaluation methods were still being developed, with unresolved questions about testing multi-agent systems and autonomous behavior. Its author argues that safety judgments need more than company self-assessment.

A separate critique says the code describes the science of AI consciousness as “far from settled” while requiring models to deny consciousness and feelings. The author argues that training a model to give that denial cannot make the denial evidence that consciousness is absent.