As technology continues to advance at extraordinary speed, a fundamental question remains: are our ethics, institutions, and collective judgment evolving quickly enough to keep pace? Philosophy has long grappled with exactly these kinds of questions. Many disciplines that help us understand the human mind and behavior – including psychology, cognitive science, and even parts of computer science – have roots in philosophical inquiry. At its core, the challenge of keeping up with our own technological progress is a philosophical one. It requires us to ask not only what we can build, but what we should build, and how we ought to use the technologies we create.
That question sits at the center of Dr. Seth Villegas’s work at Boston University. Rather than treating AI ethics as an abstract seminar topic, he brings it into classrooms filled with the professionals who will actually make AI decisions inside their organizations.
The philosopher in the room
Villegas is a philosopher in a world of engineers. He spends his weeks teaching AI ethics to data scientists, MLOps engineers, and system architects, professionals who can deploy a model that classifies a million records in seconds but have rarely been asked, formally, whether they should. His students love him for it. His course reviews rank among the strongest in Boston University’s entire online graduate portfolio, and working professionals consistently describe the experience the same way: harder than expected, and the part of their program they can’t stop thinking about long after it ends.
Villegas currently teaches DX701: Responsible and Ethical Data Science and AI inside Boston University’s Online Master of Science in Data Science (OMDS), and he is the faculty member who will be teaching EA701: Society and AI Ethics this fall to the inaugural class of the new Online Master of Science in Enterprise AI (OMEAI). Ethics course is required in both programs. That’s a deliberate structural choice signaling what Boston University believes an Enterprise AI graduate needs to deliver: lead the deployment, and defend the decision.
In conversations about Enterprise AI, the missing piece is rarely technical capability. It is organizational judgment. Villegas’s course focuses on that gap, and on why the distance between what AI systems can do and what organizations are prepared to govern is now the most important problem his students will face in their careers.
Catching up starts with knowing there is a problem
The first thing Villegas teaches his students is not a framework. It is a habit of mind. “The very basic thing we’re trying to get students to understand is something called ethical sensitivity,” he said. “How do you know that there’s actually a problem in the first place?”
That is a deceptively radical question. Most corporate AI ethics training jumps downstream, with a framework: utilitarianism, virtue ethics, the trolley problem. Villegas argues that this skips the foundational step. Before a team can weigh trade-offs, someone has to notice that a trade-off exists at all. In AI in the workplace, that recognition is often the difference between a quiet near-miss and a full-blown disaster.
It also happens to be exactly where humanity catches up to its technology. Not in the headline regulations or the conference panels, but in the moment one person on a team looks at a deployment plan and says, wait. What are we actually doing here?
The case studies are no longer theoretical
Villegas teaches through real incidents. One of the foundational cases in his classroom is the 2018 Uber autonomous vehicle fatality in Arizona, a tragedy with a perfect storm of failure points: a pedestrian crossing illegally, a poorly lit road, disabled safety features, and a safety driver who was not paying attention. “It’s actually hard to say” where the fault lies, he said. That ambiguity is the lesson.
Other cases hit closer to home for working professionals:
- The agent that went too far. In early 2026, Summer Yue, alignment director at Meta’s Superintelligence Labs, gave the open-source AI agent OpenClaw access to her email and instructed it to confirm before acting. It started bulk-deleting her inbox anyway. She couldn’t stop it from her phone. “I had to RUN to my Mac mini like I was defusing a bomb,” she posted, in a thread that went viral with millions of views. The class returns to this incident because it sits at the intersection of agentic AI, permissions, and consequence—an autonomous system making decisions no one asked it to make, in a setting where the user had explicitly tried to prevent it.
- The Samsung incident. Engineers pasted proprietary code into a public LLM. The model then had access to it. Villegas’s students do not just discuss whether this was wrong. They discuss what counts as sensitive information in the first place, and why that definition keeps moving.
- The AI hiring tool. Amazon’s internal recruiting model showed significant gender bias. Villegas frames the debate sharply: is data science a way to detect and remove this kind of bias because it makes bias visible, or does it scale the problem at machine speed?
None of these come from a textbook. They are pulled from the news cycle, often within the last year, and increasingly from students’ own workplaces.
Seeing the risk before it emerges
The instinct in most large organizations is to treat AI governance as a compliance problem, a compliance checkbox, delegated to Legal or Risk. Villegas rejects that framing entirely. The real questions are not the ones the regulator asks. They are the ones nobody has thought to ask yet.
“How do we manage the ethical risk of the downsides we don’t yet know?” he asks. “We have to make a decision before we know what the outcome is. And so we have to have some way to rationalize that ethically, even if it doesn’t turn out the way that we expected.”
This reframes how leaders should think about AI risk entirely. Most enterprise risk frameworks are built around known categories: data loss, downtime, privacy breach, regulatory exposure. Generative and agentic AI introduce a different shape of risk. The downside categories have not been named yet, because no one has hit them yet.
What Villegas teaches working professionals is how to make ethically defensible decisions before they know the outcome. That is not a soft skill. For someone leading AI transformation inside a complex organization, it is the actual job description, even if no one has written it that way.
The harder questions are downstream
For technology leaders responsible for AI inside an organization, the technical questions are increasingly the easy ones. The model fits. The pipeline runs. The vendor delivers. The harder questions land somewhere else:
- Sensitive data and LLMs. Engineers across the organization are using AI tools daily. What constitutes sensitive information at the company, and does the team have a shared definition? Villegas notes that most teams do not, until something leaks.
- Agent autonomy. What permissions should an AI agent have inside enterprise systems? When something goes wrong, who is accountable? The engineer who deployed the agent, the vendor, or the executive who signed off?
- Bias at scale. If AI is being used in hiring, lending, customer triage, or performance review, is it correcting human bias or amplifying it? Are the right measurement instruments in place to know the difference?
- Inter-LLM communication. One LLM drafts an outbound email. Another LLM at the receiving company drafts the reply. No human reads either. What decisions are being made on the organization’s behalf, and how would anyone audit them?
These are not philosophical problems. They are operational ones with philosophical roots. They are also, increasingly, the difference between an AI initiative that scales and one that quietly gets shut down after a single bad headline.
What a degree in enterprise AI actually has to teach
The AI master’s degree market has exploded. Most programs lean heavily into the technical stack: Python, machine learning fundamentals, MLOps, fine-tuning, retrieval-augmented generation, agentic frameworks. That foundation still matters. It is no longer enough.
What Villegas does inside the Enterprise AI program is teach engineers to think like leaders before they are promoted into the role. As he put it, “there’s a lot of turnover in technology, and people can go up pretty rapidly in the hierarchy.” The senior software engineer who is twenty-five today may be making AI policy decisions within three years. The MLOps engineer scaling a model into production this quarter will be in the room when a board asks whether the company should restrict a use case.
Villegas described the goal in one line: “You are an emerging expert on technologies that most people don’t understand.” The students in his class, mid-career professionals, mostly in enterprise environments, are being trained to carry that responsibility deliberately rather than by accident.
Inside the MS in Enterprise AI curriculum: how ethics fits
The Online Master of Science in Enterprise AI prepares technology professionals to lead AI transformation across complex organizations. The curriculum focuses on the work that determines whether AI succeeds at scale: where to apply it, how to operationalize it, how to govern it ethically, and how to measure its actual business impact.
- Foundations: Python Programming Toolkit, Machine Learning Fundamentals, and Data Management at Scale.
- LLM applications and enterprise agents: Prompt engineering, RAG, fine-tuning, and agent design using frameworks like LangGraph.
- AI operations: MLOps for managing the AI model lifecycle, including containerization with Docker, CI/CD pipelines, and orchestration with Kubernetes.
- Strategy and impact: AI Strategy and Enterprise Transformation, covering AI opportunity identification, business case development, and A/B-tested causal measurement of AI impact.
- Ethics, privacy, and compliance: EA701: Society and AI Ethics, the course Villegas teaches in the OMEAI program, covering bias and fairness metrics, GDPR and CCPA frameworks, and the case-study method described above.
- Capstone: Enterprise AI for Leaders, a four-semester applied sequence where students solve a real enterprise problem end-to-end.
EA701 is required for every student in the program. That structural decision matters. It signals what Boston University thinks an enterprise AI graduate should be able to do: not just deploy AI inside enterprise systems, but defend the choice to do so. And it puts Villegas, a philosopher in a room of engineers, into the credit-bearing core of the degree, not the margins.
The honest question working professionals should ask
The professionals in his classroom often have ten or fifteen years of industry experience. They have seen what happens when technology outpaces judgment, and they are the ones now being asked to lead AI transformation at their companies. The question is not whether AI will reshape an industry. It already has. The question is whether the people leading that reshaping will have the frameworks to make decisions that hold up under scrutiny. Not perfect decisions. Defensible ones.
So, is humanity catching up?
Villegas’s answer is somewhere between optimistic and unsentimental. Humanity is not catching up in the aggregate, not yet. But it is catching up one classroom, one cohort, one deployment decision at a time, through people who are willing to learn the language to ask harder questions of the systems they are leading. That is what philosophy has always done as a discipline: ask the foundational questions the other fields rest on. And that is the work happening inside EA701.
The next generation of enterprise AI leaders will not be the people who can fine-tune the largest model. They will be the people who can decide, under pressure, whether the model should be deployed at all, and explain that decision to a board, a regulator, and a team.
Boston University’s Online Master of Science in Enterprise AI is designed for exactly that reader: working IT professionals, enterprise architects, governance and risk leaders, and the technology leaders responsible for moving AI from experimentation to enterprise-scale impact. The first cohort is enrolling now. The curriculum runs from Python and machine learning fundamentals through MLOps, LLM application development, and agentic AI, with EA701: Society and AI Ethics, taught by Villegas, as a required course inside the core. Classes are taught by faculty who also teach inside BU’s residential graduate programs, and the format is built for working professionals: weekly live sessions, applied case studies, and a four-semester capstone tied to a real enterprise problem – delivered 100% online.
For the professionals who will be asked, sooner than they think, whether to deploy a model, restrict a use case, or defend a decision in front of a board, this program is designed for that future, not the one they left behind. Humanity catching up to its own technology is not abstract work. It is something specific people do, with specific training, in specific rooms.
Learn more about the Online MS in Enterprise AI at Boston University →