Prerequisite: This course is intended for students interested in technology and intending to work in the tech industry, a related nonprofit or government entity, or a tech legal practice upon graduation. We encourage students to have taken a course on a tech-related topic in their prior studies. Applicants may also cite coursework from prior institutions. Admission to the course is by permission of the instructors. Application form: https://harvard.az1.qualtrics.com/jfe/form/SV_0wcVe5J3UW9lv38. Deadline for submissions: Nov 1st 5pm ET.

Exam Type: No Exam

The recent success of large language models has led to a surge in the development of agentic artificial intelligence systems. Agentic AI – autonomous, goal-directed systems capable of planning and executing long sequences of actions on behalf of a user – are being built for various applications and are being granted increasing discretion. Some experts believe that Agentic AI systems that can pursue goals autonomously and grasp the complexities of adaptive decision-making will be integrated into daily experiences within the next two years.

Depending on how they are designed and implemented, autonomous systems can present challenges to established structures in our legal system that are designed to protect individuals from harm caused by other actors and their human and corporate agents. This course will explore the boundaries of that conflict. Students will develop concrete policies designed to ensure that society is able to take advantage of the benefits of agentic AI while mitigating when its behavior and effects inevitably go awry.

We will begin the semester with an intensive introduction, including a possible additional Friday afternoon session, intended to level-set students on technical dimensions of AI. The course will further engage with those building some of the most advanced models and tools – many of which have become part of the public imagination – to examine what gives current AI builders and leaders pause. As the semester progresses, the students will work together and with faculty to explore these topics, with the ultimate goal of producing a written term paper.

The course may conclude with a public-facing workshop on Agentic AI hosted by the Berkman Klein Center. Students and experts will present their research and ideas about what might be missing in the current legal, ethical, and policy terrain.

Note: This course will meet on average of three hours per week. The full schedule will be posted the course Canvas page closer to the start of the term.