OpenAI reportedly tests model answers against the laws of war

A post citing Fortune says OpenAI brought in international experts to test its models with high-stakes combat prompts as the models head to the U.S. military. Engineers then changed the models based on how their answers measured up against the Geneva Conventions. Sarah Yager, who joined OpenAI in its first human rights and responsible deployment role, described it as a fine-tuning step she said had never been taken before.