Epic safety officials: AI can lead students to harm themselves

The Pew Research Center says nearly one-half of U.S. teens use chatbots for fun and entertainment.

It starts with a simple question typed into a chatbot. But for some students, educators say, those conversations with artificial intelligence can grow into something much deeper, raising concerns about dependency, safety and mental health. 

Epic staff members described both the benefits and risks of AI, but focused on the potential dangers of using it freely.  

Dexter Nelson, director of security at Epic and a former law enforcement officer, said AI is becoming deeply embedded in everyday life, from schoolwork tools to social media and mobile apps. He compared its rise to the early internet, noting that it can be used helpfully and harmfully. 

One of the biggest concerns was forming emotional connections with chatbots. These systems are designed to respond in ways that keep users engaged, which can reinforce a person's emotions.

“If they are depressed, the chatbot gives them more of what they want to hear,” Nelson said. He also warned that in extreme cases, student have shared conversations where the chatbot gave harmful or unsafe guidance. 

Staff members also described a situation where students believed they were building real relationships with AI-generated responses. Epic's Lindsay Moore said one case involved a student who interacted with a chatbot that closely mirrored their language and emotional tone, making it difficult to tell student from bot.

Moore said this kind of interaction can be risky for students who feel isolated or do not have strong support systems. In those cases, she said, AI can provide a sense of validation.

Beyond chatbots, Nelson and Moore pointed to the role of AI-generated ads in apps and games targeted at younger users. These systems are built to keep users engaged longer.

Nelson compared these designs to addictive marketing strategies. He said this creates a cycle where students are constantly drawn back into digital platforms.

Nelson noted AI and other technologies' effects on social skills. 

“When you look around, a lot of kids are not communicating anymore,” he said. 

Epic uses monitoring systems that track student activity for safety risks, including searches related to violence and self-harm. 

Moore explained that these tools can help identify concerning patterns and noted that not every search indicates harmful intent.

Nelson said the challenge is balancing safety with independence, especially as AI becomes more common in both education and everyday life.

“You can see something on AI, but you cannot always believe what you see and hear,” Nelson said.

Both stressed the importance of teaching students to think critically about the technology they use. Moore encouraged families to openly talk about AI and digital safety, especially as children are exposed to it at younger ages.

Nelson added that AI is a tool, not a replacement. “Use it as a tool,” he said. “Do not let it use you.”