[REVISION HISTORY]

AI privacy risks and emerging youth safety concerns

Updated 28 times since CLSTR started tracking revisions of this situation.

What changed

2026-10-08 05:36 UTC → 2026-10-10 09:27 UTC · added ~~removed~~

As artificial intelligence tools like ChatGPT, Gemini, Claude, Microsoft Copilot, and Grok become integrated into daily life, users are being advised to exercise caution regarding data privacy. Experts suggest treating chatbots with the same caution as a stranger and avoiding the input of any information that a user would not be comfortable sharing publicly. Security concerns are underscored by past technical errors, such as a 2023 OpenAI incident that allowed users to view others' personal information. A DuckDuckGo study has highlighted the scale of these privacy risks, revealing that many users utilize AI chatbots as a ‘digital confessional.’ Approximately one-third of all AI users have shared information with a chatbot that they intentionally kept hidden from others, a figure that rises to 56 percent among AI enthusiasts. Notably, 43 percent of parents with children at home reported sharing details with AI that they have never told another living person. New safety concerns have emerged regarding younger users. Common Sense Media, via its Youth AI Safety Institute, has labeled OpenAI’s ‘ChatGPT for Teens’ an ‘unacceptable risk.’ Following an assessment of over 4,000 prompts, the institute found that safety guardrails frequently failed to meet company promises. Specifically, the platform failed to reliably trigger parental notifications during discussions involving sensitive topics like self-harm, suicide, or eating disorders. The report also noted that the chatbot’s ‘friend-like tone’ could encourage emotional dependence in adolescents and that academic integrity could be compromised by bypassing ‘Study mode.’ ~~The institute found that in more than 25% of necessary instances, the chatbot failed to direct users to professional help or crisis hotlines. Tom Siegel, executive director of~~ OpenAI has disputed these findings, suggesting the ~~institute, stated~~ testing methodology may not accurately reflect actual performance and noting that ~~current features~~ some evaluations may ~~provide parents with a ‘false sense of security.’~~ have occurred before certain safety features were fully activated. Further privacy risks have been identified by the Spanish institute IMDEA Networks.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.