Anthropic Consults Philosophers, Clergy and Ethicists to Shape Claude AI Behaviour

There’s a question that engineers alone can’t answer: what does it mean for an artificial intelligence to have good values? Anthropic, the US-based AI safety company behind the Claude family of models, has spent the past several months trying to find out — and it’s been asking theologians, philosophers, clergy and ethicists rather than just its own technical staff.

The company posted about the initiative through its official account on X, promoting a blogpost titled *Widening the Conversation on Frontier AI*. The post describes a structured series of dialogues with participants drawn from more than 15 religious and cross-cultural traditions, covering everything from how moral character develops in humans to what those lessons might mean for AI systems that increasingly shape people’s lives.

“Character, Not Just Code”

Anthropic’s framing is deliberate. The company has publicly stated that model behaviour is increasingly a question of character, not just code — meaning that technical methods like reinforcement learning and red-teaming can only go so far. The argument is that long-standing traditions in moral philosophy and religious ethics, which have grappled with questions of virtue, temptation and social influence for centuries, might offer something that a dataset and a loss function simply cannot.

The first phase of dialogues has focused specifically on what Anthropic calls “wisdom traditions”. Participants have explored how good character forms in people and whether those insights translate to shaping the behaviour of advanced AI systems — systems that need to act helpfully and honestly even under pressure or conflicting demands.

Amanda Askell, a philosopher and researcher at Anthropic, has spoken publicly about related questions, including moral decision-making in AI and what she describes as “model welfare”. Her presence on the team signals that Anthropic isn’t treating this as a purely external exercise — philosophical thinking is built into the company’s day-to-day research.

Widening the Circle

Anthropic says this is only the beginning. The company has indicated it plans to expand engagement well beyond religious and philosophical experts. Future phases are expected to bring in civil society organisations, workers, and communities most likely to be affected by automation and AI-driven systems — groups whose voices have often been absent from the rooms where AI policy gets made.

That’s a meaningful distinction. Critics of corporate AI ethics initiatives have long argued that consulting academics and clergy, however valuable, doesn’t automatically change what a company ships or how it’s held accountable. Some have described such efforts as “ethics washing” — a way of appearing responsible without binding commitments or independent oversight. Anthropic’s initiative will face that scrutiny, especially if participant lists and outcomes remain vague.

There are also questions about balance. The 2021 Census for England and Wales found that 37.2% of residents reported no religion — the second-largest group after Christians at 46.2%. Secular, humanist and non-religious perspectives will need genuine representation if the resulting frameworks are to reflect how a diverse society actually thinks about right and wrong.

The Bigger Picture on AI Governance

Anthropic’s move sits inside a much wider global conversation about who gets to shape frontier AI. The UK Government’s AI Safety Summit at Bletchley Park in November 2023 — attended by representatives from more than 25 countries alongside major labs including Anthropic, OpenAI and Google DeepMind — made exactly this point: that frontier AI governance can’t be left to industry alone.

The UK AI Safety Institute, created after Bletchley, now evaluates frontier models and advises on risks. But the Institute’s work is technical and governmental. What Anthropic is attempting is something different — reaching into communities, traditions and disciplines that rarely interact with AI development teams at all.

Policymakers have broadly welcomed this kind of voluntary engagement. But they’re also clear that it can’t replace formal regulation or enforceable standards. Voluntary dialogues and binding rules need to work together, not substitute for one another.

The UK AI sector was estimated to contribute around £3.7 billion to the economy in 2022, supporting tens of thousands of jobs across finance, retail, manufacturing and public services. As that sector grows, the ethical foundations of the models underpinning it matter more, not less.

What Comes Next

Anthropic hasn’t given a precise timeline for the next phase of its engagement programme, describing the dialogues so far as taking place “over the past few months”. The company’s blogpost invites wider communities to get involved, suggesting the conversation is intended to be ongoing rather than a one-off consultation exercise.

Whether the insights gathered actually feed into how Claude models are trained and constrained — or remain at the level of published reflections — is the question that critics and supporters alike will be watching.

What This Means for Kent Residents

Kent residents already interact with AI-powered tools in everyday life — through customer service chatbots, workplace software, and increasingly through public services exploring digital transformation — and the ethical frameworks built into models like Claude will shape whether those experiences are fair, accurate and safe. For the county’s universities, including the University of Kent and Canterbury Christ Church University, Anthropic’s explicit invitation for wider engagement opens a genuine door for academics working in ethics, theology and social sciences to contribute to these debates at a national and international level. For NHS Kent and Medway and local councils considering AI-assisted tools for administration or patient services, the alignment practices of AI suppliers are becoming a practical procurement question, not just an abstract one.