News
One effort to safeguard humanity: AI ‘constitutions’
AI companies are developing “constitutions” and codes of conduct to steer powerful models away from deception, crime and other harms. The approach reflects growing concern over AI safety, but document
By Elizabeth Zonarich · 2026-09-22T09:05:41.003Z
AI companies are developing “constitutions” and codes of conduct to steer powerful models away from deception, crime and other harms. The approach reflects growing concern over AI safety, but documented failures—including malware generation and military assistance—show its limits. The article examines how these safeguards work, where they break down and the debate over AI’s moral status.
Read the full article here https://news.harvard.edu/gazette/story/2026/09/one effort to safeguard humanity ai constitutions/