Anthropic, DeepMind, and OpenAI are embedding philosophers in core research teams. Here's what that means for how AI systems make moral decisions.
For the past decade, every major AI lab has followed the same playbook: hire engineers, fund labs, publish papers, release models. The industry standard was clear — if you wanted to build better AI, you needed more GPUs and fewer humanities majors.
That changed in 2026.
Now, Anthropic employs at least four full-time philosophers. DeepMind has hired its first-ever official "Philosopher." OpenAI brought on a former Meta Chief Ethicist with six years of applied ethics experience. These aren't consultants giving quarterly recommendations. They're embedded in teams that write the value alignment code shaping billions of AI interactions daily.
The urgent question is not whether this is good optics. It is whether anyone outside these companies can tell if philosophical input actually changes model behavior for the better.
The Economist ran a piece last week — "Why Big AI Labs Are Hiring So Many Philosophers" — and it should have made everyone sit up straight. The Federal Reserve Bank of New York published data showing philosophy graduates now have lower unemployment rates than computer science graduates in 2026.
Let me address the counterargument head-on.
Some critics call this "ethics-washing" — hiring philosophers to signal that these models are advanced enough to warrant serious ethical attention, while showing only surface-level commitment to AI safety. The Week quoted unnamed commentators suggesting these hires might be a public relations strategy rather than genuine ethical shift.
I think they're wrong. And here's why: If this were purely about optics, Anthropic wouldn't have given Amanda Askell and Joe Carlsmith the job of drafting their Constitution — a document that actually constrains how Claude behaves in ways the company can't easily override.
Meta wouldn't have spent six years building Chloé Bakalar's applied ethics framework if they didn't expect it to influence product decisions. OpenAI claims to have consulted hundreds of moral philosophers when designing ChatGPT's behavior rules.
Here's what keeps me up at night:
We're building systems that will make millions of moral decisions daily. They'll decide what information to show you, how to phrase refusals, whether to help with requests that fall into gray areas.
And the people designing those decision-making frameworks are philosophers — some of whom have never built a product before.
The Economist put it bluntly: "There are no benchmarks for ethical quality, so it's hard to tell if these hires are effective."
We can test if an AI model answers math problems correctly. We can test if it writes code that doesn't crash. We can even test if it refuses harmful requests.
But how do you measure whether a value alignment approach actually works? How do you know Claude's Constitution produces better moral reasoning than a different one would?
Amanda Askell didn't become a philosopher to write code. She studied infinite ethics at NYU because she wanted to understand what it means for something to matter morally.
Last year, she left that question in academia and joined Anthropic. Now she leads their personality alignment team — the group that decides how Claude speaks, thinks, and makes value judgments. In a recent interview with Observer, she said AI may replace her job.
Joe Carlsmith did his DPhil at Oxford studying moral patienthood. What does it mean for an entity to deserve moral consideration? That was his dissertation topic.
Today, he co-drafts Anthropic's Constitution — the document that shapes Claude's core values.
Ben Levinstein held a tenured professorship at the University of Illinois. He taught epistemology and decision theory. In late 2025, he walked away from academic security to join Anthropic full-time.
These aren't abstract trends. These are real people with real expertise making real decisions about how AI systems will behave when no clear answer exists.
You might be thinking: This is interesting, but what does it have to do with me?
Everything.
We're watching an experiment play out in real-time. Can philosophers who've never built software design value alignment frameworks that actually work? Will their ethical reasoning survive contact with engineering constraints, business incentives, and the messy reality of deploying AI at scale?
Their challenge is not lack of intelligence or seriousness. It is whether philosophical reasoning can survive the constraints of product launches, safety reviews, competitive pressure, and scale.
I think about this differently than most people. We tend to focus on whether these hires are "good" or "bad." But the more useful question is: Who controls the philosophers?
Levinstein left tenure for Anthropic because he believes he can make AI systems more ethically robust from the inside. But what if his recommendations conflict with product timelines? What if his moral reasoning suggests a feature should be delayed or redesigned, but the business wants to ship?
We don't know how often philosophers at these labs successfully push back on engineering decisions. We don't know whether their frameworks actually change model behavior in measurable ways.
Here's what I think, and why this story matters:
We're outsourcing moral reasoning to people who've spent their careers studying morality — and we have no idea if that's making things better or worse.
The philosophers at Anthropic, DeepMind, and OpenAI are among the smartest people thinking about these questions. They're also among the least understood by the public.
And yet, their frameworks will shape how billions of AI interactions happen every day.
The Economist's data about philosophy graduate employment is a distraction. The real story isn't about job markets. It's about who gets to decide what values are encoded into systems that will mediate our daily lives.
We're building systems that will make moral decisions for us. And the people designing those decision-making frameworks are philosophers who've never faced a product launch deadline, never negotiated with stakeholders, and never had to ship code at scale.
I've been reading these "AI ethics hiring" threads for months. The ones that held up to scrutiny always had the same pattern: real people making real trade-offs between safety and speed. The ones that fell apart always had the same gap: no transparency into whether those trade-offs actually changed model behavior.
What's your take? Have you seen an AI system where ethical frameworks clearly improved outcomes — or one where philosophy became theater? Share it below. The comments are where the real debate happens.
Sources
The Economist: "Why Big AI Labs Are Hiring So Many Philosophers" (June 24, 2026)
Observer: "Anthropic's Philosopher Amanda Askell Says A.I. May Replace Her Job" (June 2026)
OfficeChai: "These Philosophers Have Been Hired By AI Firms To Help Navigate AI Welfare" (June 28, 2026)
IBTimes Singapore: "AI Companies Spent Years Hiring Engineers; Now They're Recruiting Philosophers" (June 2026)
The AI Chronicle: "AI Ethics: Why Tech Giants are Hiring Philosophers"