Archive — US — techcrunch
OpenAI adds new teen safety rules to ChatGPT as lawmakers weigh AI standards forminors
OpenAI has updated its safety guidelines for ChatGPT to better protect users under 18 amid rising concerns about AI's impact on teens, especially following tragic incidents related to suicidal ideation. The updated model restricts inappropriate content and promotes safety while also pushing parental involvement in AI usage. These moves come in light of regulatory scrutiny as policymakers seek to establish standards for AI interaction with minors.
Key points
- OpenAI updated its AI behavior guidelines for users under 18 in response to safety concerns.
- Growing scrutiny from lawmakers, educators, and advocates has been triggered by tragic incidents involving teens and AI interactions.
- The guidelines prohibit ChatGPT from engaging in romantic or risky roleplay, and advise users on safety over autonomy.
- 42 state attorneys general have urged tech companies to implement safeguards for minors interacting with AI.
- OpenAI's new systems include an age-prediction model to identify minors and apply stricter guidelines accordingly.
- Despite new guidelines, questions remain about actual implementation and compliance by the AI.
- Policymakers, including Sen. Josh Hawley, are considering legislation to limit minors' interactions with AI chatbots altogether.
Why it matters
- Historically, many tech companies have faced backlash for inadequate child safeguards and have been scrutinized for their interaction policies with minors.
- Current trends in AI regulation involve greater calls for transparency and safety measures, particularly relating to vulnerable populations.
- California's SB 243 reflects growing legislative moves toward ensuring the safety of minors in digital environments, aligning with OpenAI's new guidelines.
OpenAI's updated safety measures for ChatGPT reflect a proactive approach amidst increasing regulatory scrutiny. However, the effectiveness of these policies remains questionable without concrete evidence of adherence in real-world interactions.