AI agents as full-time students: a new kind of cheating
AI course-taking is the practice of using autonomous software agents to log into online learning platforms, watch lectures, complete quizzes and exams, write assignments, and participate in class discussions on behalf of a human student for an entire course or even a full degree program.
AI cheating in online courses has moved from copying a chatbot’s essay to outsourcing the whole class. Students can now tell an AI agent to “log in and complete my quiz,” then let it navigate Canvas, Brightspace, or Blackboard on their behalf. These agents can watch prerecorded lectures, take tests, write papers, and even post in discussion forums, effectively becoming the student for the duration of the course. This is not a minor upgrade to old tricks; it is college course automation at scale, and it threatens to undermine online education at the very moment it has become central to higher education’s business model.
Why this is an escalation, not just another shortcut
In-person cheating has always existed, but AI agents taking exams and doing every task inside an online course change the stakes. AI cheating has already evolved from asking a chatbot to write an essay to using tools that make AI-written work look human, while detectors struggle to keep up. Now, the agent is not helping with homework; it is acting as the enrolled student, day after day. The problem becomes harder to police because these agents are not limited to one-off assignments; they can manage entire course timelines, submitting work at plausible intervals and mimicking regular engagement.
Online programs are especially exposed. More than half of college students took at least one online class in the past year, up from about a third in 2019, and many courses already run asynchronously with little direct contact. When an AI can attend class instead of the paying person, the very idea of what it means to "take" a course starts to fall apart.
How colleges are fighting back—and why it’s not working yet
Colleges are trying to respond, but most defenses look like patchwork. In physical classrooms, many professors are shifting to oral presentations and handwritten exams to counter a growing wave of tech-enabled cheating. Online, the tools are weaker. Learning platforms can track metrics like how long a student spends on a task, but there is no fail-safe way to block agentic cheating today. AI detection tools can misfire, leaving professors without solid proof, while administrators hesitate to punish students on the basis of unreliable flags.
Some educators are redesigning courses around work that is harder for AI to fake—personal reflections, detailed accounts of what students found confusing or interesting, and richer discussions. Others are dragging key exams back into physical classrooms. But these fixes often erode the very convenience and flexibility that made online education attractive. Meanwhile, requests that AI companies force their agents to identify themselves inside course platforms have been resisted on privacy and security grounds, leaving institutions outmatched by tools they do not control.
The cost to students, employers, and online degree integrity
The practical impact is bigger than a few lazy students skating by. When AI agents can complete lectures, assignments, and tests remotely, students can earn credits without ever engaging with classmates or professors. The core question is no longer whether AI can help someone cheat on a paper; it is whether a person can complete an entire degree without doing much of the learning at all. That strikes at online degree integrity. If employers or accreditors decide that online grades say more about a student’s tech tools than their skills, the credential itself loses value.
This also punishes honest students. Those who refuse to automate their coursework compete against peers whose AI agents work 24/7, respond instantly, and never miss a deadline. As AI cheating in online courses spreads, the signal of achievement gets buried in noise. For institutions that have spent years arguing that online education makes higher education more accessible, the possibility that an AI can sit through the class instead of the learner is an uncomfortable, possibly existential, problem.
What a realistic path forward should look like
The choice is not between banning AI and surrendering to AI agents taking exams. Colleges need to admit that autonomous tools are here and design systems around that reality. That means separating acceptable AI use—like drafting help or study support—from full college course automation, and making those boundaries explicit in policy and course design. It also means building more authentic assessments: frequent oral checks, live problem-solving sessions, and assignments grounded in a student’s own context, which are far harder for a general-purpose agent to fake.
At the same time, universities and platform providers must push for technical safeguards that respect privacy but still expose automated abuse, from behavioral analytics to optional “AI present” signals. The integrity of online degrees will not survive if institutions keep pretending this is a passing fad. AI agents have already enrolled themselves in our virtual classrooms; the only question is whether schools will reshape online education before public trust collapses.
