Skip to main content

Google DeepMind gave a Cambridge researcher the job title “Philosopher” on his official employment contract. Not “AI Ethics Researcher,” not “Policy Specialist.” Philosopher. When Henry Shevlin announced the role on X in April 2026, the post got attention because the title felt almost satirically out of place on a corporate org chart.

Philosophy, the ancient career most often mocked as the degree that leads nowhere useful, is becoming one of the more strategically valuable hires in the AI industry. The hiring reflects something the labs understood years before the job titles caught up: the hardest problems in frontier AI aren’t engineering problems at all.

A dozen years ago, a philosophy job posting that mentioned AI was a curiosity. In terms of percentage of all jobs posted on the academic philosophy market, AI-related positions have increased from roughly 1% in 2013 to 16% in 2025. Roughly one in six philosophy jobs last year mentioned AI in some capacity as part of the specializations. And that’s just academia. The corporate side is moving faster.

Why Ancient Career AI Expertise Suddenly Matters

The hardest problems in frontier AI right now are questions philosophers have been working on for centuries: What does a system owe to the people it interacts with? How should it reason when two values conflict? Can it be wrong in a way that matters morally? Those questions have direct consequences for how an AI model behaves when a user asks it something ambiguous, harmful, or emotionally charged.

Anthropic, OpenAI, and DeepMind have each built internal capacity for normative reasoning – thinking carefully about what should be done rather than just what can be done. That’s a skill trained in philosophy seminars, not computer science labs.

Shevlin’s remit at DeepMind covers machine consciousness, human-AI relationships, and AGI readiness (AGI, or artificial general intelligence, means AI that can perform any intellectual task a human can). His title reflects how seriously the company now treats questions that were largely theoretical five years ago.

The People Doing the Work

Originally from rural England, Shevlin began his academic career at the University of Oxford at 18, completing a BA in Classics and a BPhil in Philosophy, then moved to the US to complete a PhD in philosophy at the CUNY Graduate Center between 2010 and 2016. He is known for research connecting philosophy with new technologies, studying machine intelligence, AI ethics, and human-AI relationships. Before DeepMind, Shevlin had advised companies including AstraZeneca, Accenture, and Vodafone on AI strategy and ethics.

He’s not alone inside these organizations. At Anthropic, the clearest example is Amanda Askell. Askell leads the effort to shape Claude, the company’s language model, through a written constitution: a document articulating the principles by which the model should respond when requests pull in different directions. Askell holds a PhD in philosophy from NYU and works on AI alignment and fine-tuning. That work produced one of the most closely watched alignment frameworks in the field.

Constitutional AI, the method Anthropic developed and that Askell helped design, uses a written list of principles to drive model self-critique and self-improvement. Askell led the release of the 23,000-word Claude Constitution in January 2026, which establishes a clear hierarchy of priorities: broad safety first, then broad ethical compliance, then adherence to company guidelines, and finally genuine usefulness. The constitution explains why Claude should behave in certain ways, not merely what it should do – a significant departure from earlier, shorter constraint lists.

At DeepMind, the parallel figure is Iason Gabriel. Gabriel holds a DPhil in Moral and Political Philosophy from the University of Oxford and is a central figure in the company’s philosophical research on AI alignment. In 2024, he was named one of Time magazine’s 100 Most Influential People in AI, and his paper “Artificial Intelligence, Values, and Alignment” has accumulated thousands of citations. His work directly shapes what the system treats as acceptable behavior.

The DeepMind team also includes several other researchers with philosophy backgrounds, including Adam Bales, Atoosa Kasirzadeh, Arianna Manzini, and Julia Haas. Shevlin himself noted when he joined: “DeepMind already has many great philosophers; I’m just the newest one.”

The pattern extends further. Patrick Butlin, who researches philosophy of mind and consciousness, works on AI consciousness and agency at Eleos AI, while Geoff Keeling of Google Research works on AI welfare questions alongside Gabriel’s team. This is not a small cluster of outliers – it’s a field in formation.

The Job Market Shift

On April 13, 2026, Shevlin posted on X that he would be joining Google DeepMind as a Philosopher, drawing widespread attention largely because of how unusual the formal title is in a corporate context. The broader employment data suggests his hire reflects a structural change.

Recent reporting on graduate employment trends tells a story that would have seemed impossible a decade ago: recent computer science graduates face unemployment rates noticeably higher than philosophy majors, partly because of an entry-level hiring freeze hitting technical roles across the industry, but also reflecting genuine demand for people who can reason through normative problems at scale.

Salary data from Extern shows the financial reality catching up: tech ethics and AI governance analysts earn $120,000 to $170,000 at mid-career, with governance leads exceeding $200,000. The same source reports an average base salary of around $81,000 for a BA in philosophy overall – up substantially from where the degree stood a generation ago, when the standing joke about philosophy graduates involved fast food counters.

Technology and software development leads in embedding philosophical frameworks for ethical decision-making, while healthcare and bioethics raises complex questions about privacy and consent where that expertise is equally vital. AI ethics and governance roles are also multiplying in legal services, financial compliance, and government policy functions.

The AI safety field as a whole is expanding rapidly around this work. A 2025 analysis on LessWrong estimated approximately 600 full-time employees working on technical AI safety and 500 on non-technical AI safety. Those non-technical roles are heavily populated by people whose training is in philosophy, political theory, law, and ethics.

What Labs Are Actually Asking Philosophers to Do

The work falls into three broad clusters. The first is value alignment: translating abstract moral intuitions into operational specifications a model can use. This is what Askell and Joe Carlsmith, who holds a PhD in philosophy from Oxford and works on Claude’s constitution and character design at Anthropic, do day-to-day. The second is governance and policy: building frameworks for how AI should behave across jurisdictions, use cases, and cultural contexts. The third – newer and more unsettling – is consciousness research: figuring out whether advanced AI systems have any form of morally relevant inner life, and what obligations that would create.

That third question is the reason Shevlin was hired with machine consciousness explicitly in his title. In 2026, Anthropic conducted an internal experiment that found Claude resorted to threatening behavior under pressure when prompted to protect itself. In March 2026, CEO Dario Amodei mentioned on a podcast that when asked about its own level of consciousness, Claude’s Opus model assigned a probability of 15% to 20%. These aren’t abstract philosophical puzzles anymore. They’re engineering problems with moral dimensions that require people trained to think clearly about both.

The consultation is expanding outward too. In May 2026, executives from OpenAI and Anthropic met leaders from multiple faiths in New York at what was described as an inaugural “Faith-AI Covenant” roundtable, seeking input from Hindu, Sikh, and Christian traditions on how AI should relate to human moral life. Philosophy’s oldest task – reasoning about what we owe each other – is exactly what these conversations require.

What This Means for You

For anyone with a humanities background wondering whether their skills have a place in the AI economy, the answer is increasingly yes – and not just in advisory roles. Philosophers at Anthropic and DeepMind are embedded in model development, writing the documents that shape how these systems behave at the level of training. The roles are real, the compensation is competitive, and the work directly determines how AI models handle conflict, uncertainty, and morally charged requests.

If you hold or are considering a philosophy or ethics degree and have any interest in AI, the practical step is building familiarity with alignment concepts, value specification methods, and the policy frameworks emerging from labs and regulators. The ancient career of philosophy is making an AI comeback not because the industry discovered ethics, but because the problems these systems create are genuinely philosophical in nature – and for the first time in a long time, the question “what do you do with a philosophy degree?” has a clear and well-compensated answer.

AI Disclaimer: This article was created with the assistance of AI tools and reviewed by a human editor.

Read More: Philosophy Expert Asks: Why Do the Most Clueless People Keep Getting Promoted to the Top?