Microsoft Called It a Doom Loop. Your Reading Assignment Is Already Inside It.
Under oath, Microsoft’s CEO agreed the chatbot answer substitutes for the source. Your students can get most readings summarized in seconds. Assign the passage the summary leaves out.
On September 17, a summary judgment brief from the news organizations suing OpenAI and Microsoft, led by The New York Times, was unsealed in federal court in New York. Most of the coverage went to the loudest lines. The line that belongs in your next department meeting, and in your next reading assignment, is quieter. The brief quotes an internal Microsoft document: “Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time.”
The same brief says Microsoft CEO Satya Nadella agreed under oath that conversing with chatbots “has substituted … giving you the information right there on the website on the AI platform versus needing to go to the underlying source.”
Microsoft has pushed back on how the brief is being read. A company spokesperson told The Verge that the sharpest comments, made by one of its applied scientists, “reflect one employee’s individual perspective, are not a legal analysis, and do not represent the company’s views,” and that Nadella “spoke to broad principles and changes underway in how people find and consume information.” That is a fair distinction for a courtroom. For a classroom, the broad principle is the entire story.
Every reading assignment assumes the click
Every reading assignment you have ever written rests on one quiet assumption: the student opens the source. The syllabus lists the chapter. The discussion prompt asks about it. The quiz checks it. The whole chain starts with a page turned or a PDF opened, and everything you designed downstream depends on that first move actually happening.
The documents describe products built so that the first move is optional. The answer arrives on the platform, and the underlying source becomes a footnote nobody needs to visit.
Here is the uncomfortable part. The summary a student gets back is usually decent. Decent is the problem. A decent summary is enough to post a discussion reply, pass a reading check, and sound prepared for the first ten minutes of class. It fails in exactly one place, and that place happens to be the part of the reading you assigned it for.
So a rule against summaries misses the target. Students will use them. Some weeks, so will you. The sharper question is narrower and far more useful: what does a summary reliably lose? There is an answer, and it is sitting in whatever you assigned this week.
Do this while you read: the free working session below walks you through one real reading from your own course and hands you a finished class session built around the passage a summary drops. Open The Passage the Summary Drops and bring this week’s reading.
MIT answered with a book
On September 18, MIT News announced that MIT Reads, the MIT Libraries program launched in 2016, is being reimagined for the age of artificial intelligence. At its ten year mark it is turning to fiction and memoir, and President Sally Kornbluth selected Ted Chiang’s story collection Exhalation as the fall 2026 book.
The reasoning is not nostalgia. MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training urges strengthening social connection, and it cites MIT Reads as a way to “engage many more people across campus in conversation about shared norms and why community matters.” MIT Libraries Director Chris Bourg said it plainly: “Reading together also gives us the increasingly rare opportunity for both individual reflection and shared connection.”
Notice what MIT did not do. It did not write a rule against summaries. It chose a text worth reading whole and built a room around reading it together.
Tristan Harris has argued that conversation between people carries a built in reality check, the raised eyebrow and the pushback, and that chatbots are designed in a way that strips it out. A room of people who read the same page is that reality check with a syllabus attached. Most of us will never pick a campus book. Every one of us picks what happens in the first twenty minutes of class.
The passage the summary drops
A summary compresses, and what it compresses away is predictable. The qualification where the author limits their own claim. The specific example that makes an abstract point stick. The sentence where the argument turns. The voice that tells you how sure the writer actually is. Those are what a summary tends to drop, because it is built to carry the claim, not the reasoning around it.
They are also the four things you assigned the reading for. Nobody puts a chapter on a syllabus so students can repeat its thesis. We assign it so they can watch a mind work, and the working is exactly what compression removes.
Build one class session around one of those passages and the reading stops being optional. Not because you policed it. Because the thing the room is discussing is not in the summary, and students figure that out fast.
Build the session by hand this week
- Pick one reading you assign in the next two weeks. One reading, not the whole unit.
- See what the summary says. Paste the reading into the chatbot your students actually use and ask for a short summary, or write the three sentences you expect a student to walk in with.
- Find the passage it dropped. Read the original with the summary beside it. Mark the qualification, the example, the turn or the voice that went missing. Choose the one that matters most to your course.
- Write three questions only that passage can answer. If a student could answer from the summary, cut the question and write a harder one.
- Plan twenty minutes. One minute to open. Two minutes reading the passage aloud, together. Eight minutes in pairs on your three questions. Nine minutes as a full group on what the summary got wrong and why it matters.
- Close with one exit question: which sentence would you have missed if you had only read the summary?
That last question does two jobs at once. It tells you who read, and it teaches the ones who did not exactly why reading was the point. No detector, no quiz, no lecture about integrity. Just a question the summary cannot answer for them.
The free working session, The Passage the Summary Drops, runs all six steps with fields for your own reading and builds the finished session plan you can print or paste into your LMS tonight.
That is one reading
Do it once and you will see it work. Then open your syllabus. There is a reading attached to nearly every week, in every section you teach, and each reading assignment deserves the same treatment if the reading is going to stay the point of the course.
That is the job the Lesson Plan Builder does. It is stage three of the seven stage Course Creation Pathway in the Faculty Toolkit, and the toolkit describes it in its own words: “timed plans, handouts, slide outlines, video scripts.” You choose the passage and the questions, because that judgment is yours and it is the part worth protecting. The builder handles the assembly, the timed plan, the handout, the slide outline, so the fifth reading gets the same care as the first. Review first, always.
Do the first one by hand anyway. It is how you will know what good looks like when you see it.
See the tool: The Lesson Plan Builder in the Faculty Toolkit
Talk it through with other faculty: Join the free Learning Community
Planning this for a department: Book a free call
Tomorrow, before anyone opens a laptop, ask one question: which sentence in this reading would the summary have cut? The students who can answer read it. The ones who cannot are about to find out why the reading was the assignment all along.
Sources
- News Plaintiffs’ Summary Judgment Brief, public redacted version, In re OpenAI, Inc. Copyright Infringement Litigation, Case 1:25-md-03143, S.D.N.Y., filed September 17, 2026. PDF via Ars Technica
- Ashley Belanger, “Microsoft exec called AI scraping the ‘largest theft of labor in human history’,” Ars Technica, September 17, 2026. arstechnica.com
- Terrence O’Brien, “OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web,” The Verge, September 18, 2026. theverge.com
- Brigham Fay, MIT Libraries, “A new chapter for MIT Reads,” MIT News, September 18, 2026. news.mit.edu
- Report of MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training. aiandeducation.mit.edu
- Tristan Harris, interview on AI and the reality checking process. youtube.com
