AI Ethicist in 2026: The Job Inside the Company

Priya, an AI ethicist embedded with a product team at a telehealth company, spends Wednesday reviewing a new triage model. She reads the model card, runs subgroup performance across race and language, drafts an impact assessment, and pushes back on the launch timeline because the Spanish-language recall rate is 11 points below English. Friday there is a mitigation plan. Monday the team is grumpy. The launch slips two weeks. Patients get a safer product.

• Impact assessments — before deployment, not after.
• Bias and fairness audits — subgroup performance and harm.
• Policy and norms — what the company will and will not build.
• Red-team liaison — between ethics, safety, and policy teams.
• Regulatory interface — NIST AI RMF, EU AI Act, sector rules.
• Training — bring engineers and PMs up on real-world harm patterns.

• Tools like Fiddler and Credo AI — governance and observability.
• Tools like Fairlearn and IBM AIF360 — fairness toolkits.
• Model cards, system cards, data sheets — documentation standards.
• NIST AI Risk Management Framework.
• Internal review boards modeled on human-subjects IRBs.

If you want to be an AI ethicist: Interdisciplinary is the profile — CS plus philosophy, law, social science, or STS. Graduate work in ethics, policy, or HCI helps. Read the last three years of FAccT and AIES proceedings. Publish or blog honestly about real cases. Industry hires from academia, consultancies, civil society, and sometimes from inside product teams. Build relationships with engineers — the work only moves if they trust you.