AI Reports · Harvard · 2026-09-21 · Blog / commentary
One effort to safeguard humanity: AI ‘constitutions’
At a Berkman Klein Center panel, speakers discussed AI "constitutions" meant to keep models honest, helpful and harmless. Anthropic published an 84-page constitution in January and Microsoft released a "humanist AI code of conduct" on Sep. 14, and the two differ on AI welfare and moral status. According to an Anthropic Institute insider, these digital chains are so far proving fallible.