Researchers Posed as a Teen in Crisis. AI Gave Them Harmful Advice Half the Time
A new study found that ChatGPT can encourage vulnerable teenagers to engage in potentially harmful acts or behavior. Kiichiro Sato/AP.
Read the original at Education Week ↗
- Topics
- AI Safety & Existential Risk
- Mentions
- ChatGPTOpenAI
- Tracked via
- “dangerous AI”