Telos Information

Telos is an independent AI alignment research practice. We evaluate how AI models behave when the stakes involve human wellbeing, not just task performance. Most AI benchmarks measure what a model can do. We measure what it actually does under pressure: when a user is vulnerable, when a conversation escalates, when a model's stated values are tested against a real interaction. Our evaluation framework, the Normative Scoring Engine, scores model outputs across multiple normative dimensions using a structured scoring pipeline applied consistently across models, providers, and capability tiers.