Philadelphia AI Safety
Deepfakes in the Hallway: Philadelphia Schools Respond to Harmful Student AI
A Philadelphia-focused response to AI impersonation, humiliating images, rumor networks, consent, reporting, and digital citizenship.
2026-08-26 · 13 min read
Easy-read summary
The big idea for Philadelphia, Pennsylvania schools
Philadelphia school leaders may first hear about student AI misuse through an essay, but the most urgent incident could arrive as a fake image or cloned voice moving through group chats. Generative tools can place a classmate in a humiliating scene, imitate a teacher, manufacture a threat, or produce a screenshot that appears authentic. Calling the content artificial does not make the fear, embarrassment, or reputational damage artificial. Middle and high schools need a safety plan that stands beside their academic-integrity policy. It should explain consent, reporting, evidence preservation, support for targeted people, and consequences for creation or distribution. The central message is direct: access to someone's public photo or voice does not grant permission to turn that identity into deceptive content.
Teach students to disclose AI help when a teacher or assignment requires it.
Use AI as a coach for questions, practice, feedback, and revision—not as a ghostwriter.
Verify important AI claims with reliable sources before sharing or submitting them.
Protect student privacy by avoiding personal, family, medical, or school-identifying details in prompts.
Section 2
Artificial media creates real harm
Consent education should come before a creative AI assignment. Students may understand that touching another person's property requires permission while failing to see a face, voice, or personal story as something deserving similar respect. Philadelphia teachers can present scenarios involving parody, campaign videos, historical reenactments, memorial content, and jokes between friends. Students ask who is represented, whether permission exists, whether viewers could be deceived, and who might be harmed. Even disclosed synthetic media can violate dignity. Classroom projects should use licensed materials, fictional identities, or students who have given specific informed permission and can withdraw it. This practice builds a boundary students can carry into social spaces where the creation tool offers no ethical pause.
Teens reported using generative AI for school assignments in Common Sense Media research.
RAND reported student AI homework use among middle school grades and up by December 2025.
College Board reported high school students using generative AI for schoolwork.
Section 3
Teach consent before teaching creation
A reporting system must be simple enough to use under stress. Students should know one adult, one online route, and one emergency route. Reports should allow links or screenshots without requiring the student to reshare harmful files widely. Staff need a response checklist: secure evidence, assess immediate safety, contact designated leaders, support the person targeted, limit further circulation, and communicate without repeating the content. Philadelphia schools can publicize the route on student portals, classroom posters, and family materials. Anonymous tips may help surface incidents, but students should also understand when a direct conversation is necessary to provide support. Speed matters because synthetic media can cross classrooms and neighborhoods long before a traditional investigation begins.
Section 4
Build a rapid reporting pathway
Bystanders are often the largest audience and the best chance to interrupt harm. A student who forwards a deepfake ‘so you know’ increases its reach, even without creating it. Schools can teach a four-step response: do not repost, capture only what is needed for a report, tell a trusted adult, and check on the person targeted. Students should avoid public arguments that feed attention or expose the victim further. Peer leaders, athletes, performers, and club officers can model this standard because their online choices shape group norms. A campaign built around ‘pause the spread’ is more actionable than a generic warning to be kind. It gives students something useful to do in the seconds when sensational content appears.
Section 5
Stop the forwarding chain
Homework substitution and synthetic harassment require different procedures. An AI-written paragraph primarily concerns evidence of learning; a sexualized image, impersonated threat, or targeted rumor concerns safety and dignity. A single ‘AI violation’ category risks trivializing one case or over-penalizing another. Philadelphia administrators can create a decision tree separating academic misuse, privacy exposure, harmful media, and credible safety threats. Each branch identifies relevant staff and policy. The same incident may involve several branches—for example, fabricated research about a classmate used in a presentation. Clear classification improves consistency and helps families understand why two uses of generative technology led to different responses. Technology is the method; intent, impact, and learning context determine the response.
Student prompt examples
- “Coach me, but do not complete the assignment for me.”
- “Ask me three questions before giving suggestions.”
- “Show what might be wrong or missing in this answer.”
Section 6
Separate safety cases from homework cases
Media literacy lessons can reduce both creation and circulation. Students can examine benign synthetic samples and look for provenance, original uploads, inconsistent context, missing attribution, and trustworthy corroboration. They should learn that visual oddities are not a reliable long-term detection method because tools improve. Better questions are: Who posted this first? What evidence supports the event? Has a credible source confirmed it? What does the person represented say? Teachers can use reverse-image searching where available and discuss content credentials without promising a perfect technical solution. The habit is to pause before belief and pause again before sharing. That discipline supports civic education, history, journalism, health, and everyday peer relationships.
Section 7
Practice media verification in class
After an incident, discipline alone may not repair the community. The targeted student may need counseling, schedule support, help with takedown requests, and reassurance that adults will not repeatedly expose the material. The student responsible should face appropriate consequences and learn about consent, impact, and digital permanence. Restorative processes should occur only when safe and desired by the harmed person; they must never require a victim to educate the creator. Schools can communicate the behavioral standard to the wider community without identifying students or replaying details. A review afterward should ask where reporting slowed, how the content spread, and what supervision or education should change. That review converts a painful event into stronger prevention.
Section 8
Repair trust after an incident
Jim Jordan's Philadelphia AI school assembly places character at the center of digital power. He speaks directly about deepfakes, privacy, reputation, and the moment a student's choice can either protect or expose another person. His twenty years on stages with young people help him address serious harm without losing the room, and his JimmyAI experience provides credibility about what the tools can actually do. Schools can pair the keynote with consent scenarios, the rapid reporting pathway, and bystander practice. Jim's featured message is not that students should fear technology. It is that increased creative power demands increased human responsibility. That principle can unite neighborhood, charter, career-technical, and independent school communities across Philadelphia.
Action checklist
Use this after the assembly
- ●Create green, yellow, and red examples of acceptable AI help.
- ●Give students simple prompt language such as: coach me, do not complete it.
- ●Ask students to cite or explain how AI was used when appropriate.
- ●Revisit AI expectations during major writing, research, coding, and media projects.
Sources and citations