Impact on campus
Opportunities and risks from AI agents in higher education. Every item here involves an agent that takes action: browsing, submitting, changing records, or running workflows.
Each item is labelled by evidence type. A vendor claim or institution-reported result is not the same as an independent evaluation, and observations are not confirmed findings. None of these items yet reports measured learning outcomes.
Opportunities
Agents campuses chose to deploy, with early results.
- OpportunitiesInstitution-reportedPontificia Universidad Católica de Chile · 2025
AyudantIA: UC Chile’s project bringing AI into the classroom
Faculty build their own course agents from validated materials. The university reports 170+ agents and access for 10,000+ students. Learning outcomes are still being measured.
- OpportunitiesVendor claimInstructure (press release) · Mar 12, 2026
Instructure delivers on its agentic AI promise with IgniteAI Agent
A vendor agent inside Canvas that builds modules, adjusts due dates, and runs course workflows from one prompt. Opt-in at institution, department, and course level.
- OpportunitiesIndependent analysisPhil Hill, On EdTech · Mar 26, 2026
Instructure releases the IgniteAI Agent: now the real test begins
An independent read on the launch. Campus adoption is cautious, and it is too early to judge impact.
- OpportunitiesResearch benchmarkCIKM 2026 AnalytiCup / University of Kassel · 2026
Agentic AI in university administration (UniAgent shared task)
Researchers test agents on real, anonymized administrative cases. Agents are scored on whether they reach the same decision as staff, and must log their tool use and reasoning.
Risks
Agents acting in campus systems without the institution’s consent.
- RisksNews reportingInside Higher Ed · Feb 26, 2026
Agentic AI can complete whole courses for students. Now what?
A commercial agent logs into Canvas daily, watches lectures, posts to discussions, and submits work. Faculty warn it undermines the value of online credentials.
- RisksDocumented testColorado State University, TILT · Nov 2025
New AI browsers can complete Canvas quizzes. Here’s how to secure them
The academic integrity office recorded an agentic browser scoring 5/5 on a Canvas quiz from a simple prompt, and recommends a lockdown browser.
- RisksFaculty observationsCalifornia Virtual Campus · Nov 2025
Safeguarding academic integrity in the age of agentic AI
Community college faculty report agents logging in with student credentials and completing coursework while posing as ordinary browsers. Also flags security and privacy risks in student portals.
- RisksCompiled demonstrationsAnna Mills · 2026
Agentic AI: considerations for educators
Collects demonstrations of browser agents taking LMS quizzes, and evidence that LMS logs cannot reliably detect them.
- RisksAuthor’s observationsJoe Sabado (this site’s author) · 2025
AI wearables and agentic browsers: integrity, equity, and literacy on campus
Campus accounts of an agentic browser signing in past Duo two-factor, and of one grading assignments and writing feedback under a faculty member’s name without being asked.
Institutional responses
Guidance and policy written specifically for agents.
- Institutional responsesPolicy statementModern Language Association · Oct 2025
Statement on educational technologies and AI agents
Calls on lawmakers, LMS vendors, and AI companies to give institutions the ability to block agents, and for faculty to have a say in procurement.
- Institutional responsesInstitutional guidanceUniversity of Missouri System · 2025
AI agents: guidelines and concerns
Faculty guidance on agents in Canvas. Warns agents can read restricted data such as class rosters, and suggests proctoring and in-class work.
Summaries are this guide’s own. Follow the links for the full articles. Inclusion is not an endorsement.