Harvard has been putting courses online since October 2012, when HarvardX launched on edX with CS50x and a public health statistics course and more than 100,000 people signed up in the first weeks. Since then the picture has got more complicated. HarvardX is the free-to-audit MOOC layer that still sits on edX. Harvard Online, launched in January 2021 with a course called Data Science Ready, is the university's own paid short-course brand, originally delivered on the Harvard Business School Online platform.
HBS Online and the executive programmes sit above both. Professional and Lifelong Learning, at pll.harvard.edu, is the catalogue that stitches all of it together, and it is the page most people reach when they search for a Harvard AI course. That catalogue is worth reading carefully because the prices tell you which layer you are in. Filter it to artificial intelligence and you get 45 results. At the bottom, free with an asterisk, are the HarvardX courses: CS50's Introduction to Artificial Intelligence with Python at 7 weeks and 10 to 30 hours a week, Machine Learning and AI with Python at 4 to 5 hours a week, Introduction to Data Science with Python taught by Pavlos Protopapas, Using Python for Research, Data Science: Building Machine Learning Models, and the entire CS50 family.
The asterisk means that watching everything is free, but graded assignments and the certificate require a verified upgrade, which is $299 for CS50 AI and the Python ML course and $149 to $219 for the data science courses. In the middle, at $495 to $1,600, are the Harvard Online courses: the three AI in Medicine courses at $495 each, Agentic AI Foundations and Reinventing the Organization with AI at $595 for four weeks, Data Science and AI Principles at $1,200 for five weeks with live faculty sessions, and Digital Health at $1,600. Then HBS Online with AI Essentials for Business and AI for Leaders at $1,949, and at the top the in-person executive seminars at $5,100 to $11,700. The free layer is the reason to care. CS50 is the best taught introductory computer science sequence I know of anywhere, and CS50 AI is a proper course with search, knowledge, optimisation, learning, neural networks and language, graded by an autograder that does not care whether you paid.
I reviewed it separately on this site. Machine Learning and AI with Python is an intermediate course on decision trees, bagging, random forests and gradient boosting, which is a sensible slice of what working data scientists actually use. In April 2026 Harvard announced six more free courses, including Data Science: Machine Learning and Introduction to Data Science with Python, and the catalogue has a 20 percent promotion code on certificates running through the autumn, which tells you edX is working harder to convert free learners than it used to. The paid layer is where my enthusiasm cools. The Harvard Online business courses are taught by real Harvard faculty and the production is good, but $595 for four weeks of asynchronous video on agentic AI, or $1,949 for AI Essentials for Business, is a price set for corporate learning budgets.
They are fine if someone else is paying and you want the name on a certificate. If you are paying yourself, the free HarvardX courses teach more about how these systems work, and the business courses do not teach you to build anything. The data science side deserves a caveat. The nine-course Professional Certificate in Data Science by Rafael Irizarry is one of the better statistics-first introductions online, and it is free to audit, but it is taught in R with tidyverse and was recorded years ago. If your goal is a job in machine learning rather than a grounding in statistics, the Python courses are the better fit, and the certificate alone will not move a hiring manager. There is also a structural point that the Harvard name hides.
HarvardX courses run on edX, and edX has not been Harvard's since 2021, when Harvard and MIT sold it to 2U for about $800 million. 2U filed for Chapter 11 in July 2024 and emerged that September as a private company owned by its former lenders. The courses carried on and Harvard told its faculty students would not be affected, but the platform that hosts the free courses is now a leveraged education company rather than a university nonprofit. MIT responded by pulling its courses onto its own MIT Learn platform.
Harvard has so far stayed on edX while building Harvard Online for the paid tier. That is worth knowing before you pay for a multi-course certificate whose completion you expect to take a year. My own recommendation is simple. Audit CS50 AI and Machine Learning and AI with Python for nothing, do the assignments properly, and pay the $299 only if a certificate solves a real problem for you. Use the catalogue's price filter to stay in the free tier unless an employer is paying.
Harvard's free online teaching is some of the best there is; its paid online teaching is mostly a brand licence with good lighting.