Anthropic has updated Claude's usage policy to prohibit sustained, excessively cruel behavior toward its AI models, alongside tighter election misinformation rules.