OpenAI updated ChatGPT with a memory system that tracks personal context across conversations. While the update aims to make research easier, it raises privacy questions for students using the tool for schoolwork. Educators and parents need to know how this tracking affects data safety.
What Happened
Under the latest OpenAI rollout, ChatGPT Plus and Pro users in the United States have a new memory system. The update replaces individual saved memories with a "dreaming" architecture to track context across long projects. Data from technical evaluations of the Dreaming V3 system shows that ChatGPT’s factual recall success rose from 41.5% in 2024 to 82.8% in 2026, and preference adherence reached 71.3%. The update adds a memory summary page and memory sources to trace where data comes from. It also includes a Gmail integration that reads user inboxes to manage schedules. OpenAI plans to roll these features out to free users soon, though Reference Chat History is currently unavailable to Enterprise and Edu customers.
The Bigger Picture
Many educators worry about persistent digital profiling. Assistant Professor Rakibul Hasan warned in an Arizona State University report that school platforms generate large digital footprints. This risks creating a surveillance system that infers private personal traits from basic user behavior. This tracking can damage student trust. An Information Research study found that students often withdraw from or resist using mandatory digital systems when they do not understand how the algorithms work.
Security risks increase when students connect personal accounts. A cybersecurity analysis shows that linking Gmail to an AI model exposes private messages and files. This leaves accounts vulnerable to "ZombieAgent" attacks, which run hidden commands to steal data. The GitLab Advisory Database also documented security failures where connected AI tools sent outbound emails without user consent.
These security risks have caused academic backlash. Over 700 students and staff signed a University of Colorado Boulder protest letter against their school's partnership with OpenAI, citing concerns about data privacy. An Ogun Security report showed how default settings in ChatGPT Edu environments silently exposed student research data and repository names to peers at other schools.
What This Means for Families
For parents and educators, the main concern is that early mistakes might permanently shape a student's AI profile. Because AI memories persist, a student who struggles with a math concept early on might find the tool continues to treat them as behind, even after they master the material. Students also tend to agree to terms of service quickly or link accounts to save time. By doing this, they can share their schedules and school work with third-party tracking systems. As we previously reported on classroom screen time, tech convenience often comes with a hidden cost to student privacy.
What You Can Do
To protect student data, parents and educators can take several direct actions. First, turn off the memory tracking feature by going to Settings, then Memory. You can also use the summary page to delete specific details the tool has saved. Second, keep personal accounts disconnected. Avoid linking Google Drive or Gmail to any account used for homework. Finally, teach students to use the "Temporary Chat" feature so their school drafts and research do not build a permanent profile.