Your Role in AI Safety

AI safety and governance are not problems for someone else to solve. The decisions being made right now — in AI labs, in legislatures, in standards bodies, in courts — will shape the trajectory of this technology for decades. The people making those decisions include researchers in their late twenties who started thinking seriously about AI safety when they were your age, policy analysts who studied computer science and philosophy in college, engineers who asked uncomfortable questions at their first internship, and journalists who learned enough technical depth to hold companies accountable. There are concrete paths from where you are now to meaningful contribution — and they do not all require becoming a machine learning researcher.

Skills That Transfer Directly

Several clusters of skills have high value across virtually every pathway into AI safety and governance work. Mathematics and statistics: linear algebra, probability, and calculus are the foundations of understanding how AI systems work. You do not need a PhD to reason rigorously about bias metrics, confidence intervals in safety evaluations, or the mathematical properties of alignment algorithms. Strong quantitative foundations open both technical research paths and technically informed policy paths. Programming and systems thinking: the ability to read and write code — particularly Python, the dominant language in AI development — lets you engage with AI systems not as a user but as someone who understands their construction. Even a working knowledge of how to fine-tune a language model, run a bias evaluation, or read a model card puts you substantially ahead of most policymakers and journalists in your ability to engage critically. Philosophy and ethics: formal training in logic, epistemology, and ethics is directly relevant to alignment research, AI governance analysis, and policy work. Philosophy departments at universities offer exactly this training, and it is more relevant to AI safety than most people recognize. Writing and communication: much AI safety work involves translating between technical communities and policy communities. The researcher who can write a compelling policy brief, the engineer who can testify accessibly before a legislature, the journalist who can explain an interpretability paper to the public — these people are in short supply and high demand. Law and policy: legal and regulatory frameworks are the operational instruments of AI governance. Understanding how legislatures work, how regulatory agencies operate, how international law functions, and how litigation shapes industry behavior is directly applicable.

Near-term actions while in high school: Read primary research. The AI safety field produces publicly accessible research at arXiv (arxiv.org), on organization websites (Anthropic, DeepMind, OpenAI, ARC, MIRI, CHAI), and in accessible blog posts. Reading even one paper per month and engaging seriously with its arguments builds genuine expertise faster than any course. Engage with AI systems technically, not just as a user. Free tools — Google Colab, Hugging Face, PyTorch tutorials — allow you to build, fine-tune, and evaluate small AI models. Hands-on experience with what models can and cannot do is irreplaceable. Write about it. A blog, a school publication, or a submission to a science or social-science competition that applies serious analysis to an AI safety or governance question demonstrates capability in a portable, verifiable form. Seek programs and communities. The Center for AI Safety runs a course on AI safety fundamentals. AI Safety Camp (hosted online) brings together students for collaborative safety research projects. BlueDot Impact runs AI safety fundamentals courses. 80,000 Hours, an organization focused on high-impact careers, has extensive resources specifically on AI safety as a career path, including honest assessments of what bottlenecks the field actually has and where students are most needed.

Post-secondary paths: Technical research path: undergraduate studies in computer science, mathematics, or statistics, with coursework or independent study in machine learning. Research experience through summer programs (REUs in the US, similar programs elsewhere), undergraduate research with a professor, or internships at AI safety organizations. Graduate study in machine learning with an AI safety focus at institutions including MIT, Stanford, UC Berkeley, CMU, Oxford, and Cambridge — all of which have faculty working on alignment, interpretability, or AI safety-adjacent problems. Policy and governance path: undergraduate studies in public policy, political science, law, or a combination with computer science. The AI policy field has significant demand for people who understand both the technical reality of AI systems and the mechanics of governance. Organizations including the Center for Security and Emerging Technology (CSET) at Georgetown, the AI Now Institute, the Future of Humanity Institute (Oxford), the Centre for the Governance of AI, and government digital services offer internship and fellowship pathways. Law path: law school followed by specialization in technology law, with particular focus on AI liability, intellectual property, privacy, and regulatory law. As AI governance frameworks mature, the legal industry will need large numbers of attorneys who understand AI systems at a technical level. Journalism and communication path: studying journalism, communication, or public policy while developing technical literacy sufficient to evaluate and explain AI research. The field of technology journalism that covers AI with genuine technical depth is small and its output is highly influential.