Superpower Daily
Amber Raises €7M to Build AI That Organizes SME Data Before It...OpenAI Pauses Frontier Training After Sandbox Breach, Warns of...Blitzy and XBOW Push AI Beyond Short Tasks Toward Continuous En...EverestLabs Puts AI Agents Near Recycling Controls, but Keeps O...Lexis+ with Protégé Adds Agents, Context and Office Files to Le...Seismic Completes Highspot Merger as Its Survey Finds 9% Embed...Amber Raises €7M to Build AI That Organizes SME Data Before It...OpenAI Pauses Frontier Training After Sandbox Breach, Warns of...Blitzy and XBOW Push AI Beyond Short Tasks Toward Continuous En...EverestLabs Puts AI Agents Near Recycling Controls, but Keeps O...Lexis+ with Protégé Adds Agents, Context and Office Files to Le...Seismic Completes Highspot Merger as Its Survey Finds 9% Embed...
Culturepublished

Cheshire Academy Uses Traffic-Light Labels to Set AI Boundaries in Class

The Connecticut school is testing a flexible model: teach AI’s limits, let teachers set task-specific boundaries, and make students examine the tool’s effect on their work.

By 2 min read
Cheshire Academy Uses Traffic-Light Labels to Set AI Boundaries in Class
Sign in to save and react

Listen to this story

The audio brief

About 1:24
0:001:24
Read transcript
At Cheshire Academy, every assignment can now carry a traffic-light rule for generative AI: green means use is allowed, red means it is prohibited, and yellow means only the tools a teacher names are permitted. That might mean spell-check is fine, while messaging a chatbot is not. The Connecticut boarding and day school is testing this model with about 400 students in grades nine through twelve. The key distinction is that AI permission belongs to the task, not automatically to the whole course—and teachers are not required to adopt it. The school uses MagicSchool alongside ChatGPT and Perplexity, while training staff to write prompts and recognize incorrect or biased outputs. In some classes, students are asked to challenge the machine rather than accept its answer. In French, for example, students have a large language model edit their work, then judge which changes improve accuracy and which erase their own voice. Teachers use AI for planning, lesson materials, and grading rubrics, but the school is holding back from AI-generated student feedback because of concerns about quality, personalization, and privacy. A Student AI Council is also leading discussions and media projects about responsible use. The open question is whether these assignment-level boundaries and critique exercises will actually change how students use AI—something the pilot has not yet established.

Story brief

3 key points

Cheshire Academy is piloting an assignment-level framework that lets teachers mark generative-AI use green, yellow, or red, while leaving adoption optional for instructors. The Connecticut school serves roughly 400 students in grades 9–12 and mixes MagicSchool with ChatGPT and Perplexity rather than standardizing on one product. It trains staff to evaluate errors and bias, uses AI for planning and selected student...

  1. 01

    Yellow can allow spell-check while banning chatbot messaging, making permissions task-specific rather than course-wide.

  2. 02

    French students critique model edits for accuracy and whether they erase the student’s voice.

  3. 03

    Technology coordinator George Aiello says most instructors use AI, but the school does not require it.

At Cheshire Academy, an assignment can now carry a simple instruction for generative AI: green means it is allowed, red means it is banned, and yellow allows only the tools a teacher specifies. The system gives teachers a way to distinguish useful assistance from work students must do themselves, without requiring every instructor to use AI.

The approach is being tested at a private Connecticut boarding and day school with about 400 students in grades 9 through 12. Rather than mandate one product, Cheshire Academy uses MagicSchool alongside general-purpose chatbots including ChatGPT and Perplexity.

Rules attached to the work

The labels make the permission decision specific to an assignment. Under yellow, a teacher can permit a narrow aid such as spell-check while barring students from messaging a chatbot. Green permits AI use; red prohibits it.

That flexibility sits alongside staff training on prompt writing and on the technology’s failure modes, including incorrect and biased outputs. The school does not require instructors to use AI, though technology coordinator George Aiello said the vast majority use it in some form.

Teaching judgment, not just prompting

Some of the school’s classroom exercises put the model’s output under scrutiny rather than treating it as a finished answer. French teacher Miriam Przybyla-Baum asks students to have a large language model edit their work, then decide which changes are accurate and which strip away the student’s voice.

Where the school is drawing the line

  • Teachers commonly use generative AI to prepare class materials, including lesson plans and grading rubrics.
  • The school has not adopted AI to provide student feedback, citing concerns about quality, personalization, and privacy.
  • A pilot Student AI Council has students create media and lead discussions about healthy AI use.

A tool choice remains a teacher choice

MagicSchool is one option in the school’s mix. It has free and paid versions; an individual plan costs just under $100 a year for unlimited access and complete records. Teachers can also use general-purpose chatbots for administrative work.

The more consequential boundary is not which generator a teacher selects, but where its output can enter the learning process. Cheshire Academy permits experimentation in planning and selected student tasks while holding back the more personal job of feedback. Whether its labels and reflection exercises change how students use AI is not yet established.

Sources

  1. technologyreview.comHow to encourage smarter AI use in the classroom

Related coverage

Continue reading