AI Labs Recruit Philosophers to Shape Model Behavior and Ethics

Google DeepMind and Anthropic have built in-house philosophy teams to guide AI development. WIRED counts at least 10 philosophers at DeepMind and four at Anthropic, though both labs declined to disclose exact numbers citing company policy.

Philosophers at these labs focus on immediate risks—fairness, misinformation, malicious misuse, errant agents—rather than speculative questions about AI consciousness. At Anthropic, philosopher Amanda Askell drafted Claude's "constitution," a document that stipulates how the model should behave and what values it should uphold; at DeepMind, Julia Haas recently coauthored a Nature paper proposing a framework for testing whether language models exhibit moral competence.

Some academics worry this arrangement risks "ethics-washing"—using philosopher hires to signal commitment to safety while constraining their research to competitive business goals. Edward Harcourt, professor of philosophy at the University of Oxford, argues that "if you're a philosopher working for a big tech company, your problem space is delimited."