Six AI-in-education stories from the last few days: a randomised trial where an AI tutor lowered grades, how Utah schools are rolling out Gemini, a new Colorado honour for campuses, and a landmark student-data standard.
1. UMD trial: course-integrated AI tutor linked to lower grades
Source: The Syllabus (reporting a University of Maryland working paper) · Oct 6, 2026
A randomised trial across 30 instructors' courses and 2,379 students gave some sections a GPT-4o-based study assistant. According to the newsletter summary of the working paper, final grades fell by 0.27 to 0.37 standard deviations, and participation in instructor-designed activities dropped sharply.
First-generation students reportedly lost more than twice as much. Only about 15% of students with access used the tool, and nearly three-quarters of requests asked for explanations or solutions, while under 1% asked for feedback on their own work.
The lesson: the tool matters less than how students are told to use it.
2. Utah rolls out Google Gemini to schools, district by district
Source: KSL.com · Oct 7–8, 2026
Utah education officials made Google Gemini available to public schools for 2026-27, but each district decides how to use it. Canyons School District has switched it on for teachers, staff and high school students, with younger pupils excluded.
Canyons uses a "stoplight protocol": red means no AI, yellow means AI only as the teacher specifies, green means approved use. Provo City School District has told parents chatbots are off on student Chromebooks while its policy is written, and the state is still drafting guidance.
Clear, simple rules beat blanket bans or free-for-alls.
3. Colorado creates an "AI Engaged Campus" designation
Source: Colorado Public Radio · Oct 8, 2026
The Colorado Department of Higher Education has launched a state honour for colleges that take a deliberate approach to AI, with intentional policies and trained staff and faculty. The first six designees include the University of Denver, Colorado School of Mines and the University of Northern Colorado.
The designation carries no funding. Officials say it is meant to help prospective students judge which institutions to trust, not to push AI use, which they acknowledge is already widespread.
Expect other states to copy the idea.
4. UAE education minister: AI cannot replace human interaction
Source: The National · Oct 6–7, 2026
Speaking at the Education First Forum in Abu Dhabi, Education Minister Sarah Al Amiri said technology and AI are tools to be used consciously, and that no educational transformation can happen without highly capable teachers.
UAE University's vice chancellor said every student will need to complete a compulsory AI course before graduating, while Emirates College for Advanced Education described using AI for personalised feedback in a human-centred way.
A useful counterweight to the "AI replaces teachers" headlines.
5. Microsoft, AFT and UFT agree an enforceable AI privacy standard
Source: The Syllabus · Sept–Oct 2026 coverage
Reported as announced in September and taking effect on Nov. 1, the standard bars using student or educator data to train AI models and bans AI companions and student monitoring. It also requires a human in the loop for AI-driven decisions.
Districts can pause or end a deployment if Microsoft breaches the terms, and talks with Anthropic and OpenAI to adopt similar terms are reportedly ongoing.
Watch whether other AI vendors sign up.
6. Pearson and McGraw Hill keep buying AI assessment and classroom tools
Source: The Syllabus (citing company announcements) · Sept 29, 2026
Pearson agreed to acquire Workera, an AI-native skills assessment platform, and earlier bought Internet Testing Systems, a remote testing and proctoring company. McGraw Hill meanwhile bought classroom audio analysis firm TeachFX, shortly after buying Teachally. Deal terms were not disclosed.
The pattern points to big publishers building AI-driven assessment and teaching analytics into their core products.
Consolidation can mean fewer, bigger vendors setting the rules.
The AI tutor study that lowered grades
A 2,379-student randomised trial is the most debate-worthy story today: parents, teachers and students all want to know whether AI tutors help or hurt, and the numbers cut against the hype.
An AI tutor was given to thousands of students, and their grades went down. Here is what the researchers think went wrong.
Join the newsletter