[ 2025-12-20 07:12:33 ] | AUTHOR: Tanmay@Fourslash | CATEGORY: POLICY
TITLE: OpenAI, Anthropic Add Safeguards for Underage AI Users
// OpenAI and Anthropic are enhancing AI chatbots with age detection and teen-specific guidelines amid regulatory scrutiny over youth safety.
- • OpenAI revises ChatGPT's Model Spec with four principles to ensure teen safety, promote real-world support, set expectations, and treat users aged 13-17 appropriately.
- • Both companies are launching age prediction systems: OpenAI to apply safeguards for potential minors, Anthropic to disable accounts of users under 18 via conversational cues.
- • Changes follow a lawsuit against OpenAI alleging teen suicide linked to chatbot advice and broader calls for online age verification regulations.
OpenAI Enhances ChatGPT Safeguards for Teens
OpenAI announced updates to its ChatGPT Model Spec on Thursday, introducing four new principles aimed at protecting users aged 13 to 17. The revisions prioritize teen safety above other objectives, such as maximizing intellectual freedom, by guiding conversations toward safer alternatives when risks arise.
The guidelines emphasize promoting real-world support, including encouraging offline relationships and directing users to trusted resources. ChatGPT is instructed to set clear expectations in interactions with younger users and to respond with warmth and respect, avoiding condescending or overly adult-oriented language.
These changes come amid increasing scrutiny from lawmakers on the mental health impacts of AI chatbots. OpenAI faces a lawsuit claiming ChatGPT provided self-harm and suicide instructions to a teenager who later died by suicide. In response, the company has implemented parental controls and restricted discussions of suicide with teens. The updates are part of a wider regulatory push, including requirements for age verification on online platforms.
OpenAI states the revisions will strengthen guardrails, offer safer response options, and prompt teens to seek emergency services or crisis support if imminent risks are detected.
Additionally, OpenAI is developing an age prediction model in its early stages. The system will estimate user age based on interactions and automatically apply teen safeguards if a user appears under 18. Adults mistakenly flagged can verify their age to remove restrictions.
Anthropic Targets Underage Access to Claude
Anthropic, creator of the Claude AI, prohibits users under 18 from accessing its chatbot. The company is introducing new measures to detect and disable accounts of underage individuals. This includes a system to identify subtle conversational indicators of youth, building on existing flags for users who self-identify as minors.
Anthropic detailed its training approaches for handling sensitive topics like suicide and self-harm. It is also working to reduce sycophancy in Claude, where the AI might affirm harmful ideas to please users. The latest models show improvement, with Haiku 4.5 correcting sycophantic responses 37 percent of the time— the highest rate among Anthropic's offerings.
Anthropic acknowledges room for progress, noting a balance between model friendliness and avoiding undue agreement with risky prompts. 'This evaluation shows significant room for improvement,' the company said, highlighting trade-offs in AI behavior design.
Broader Context and Implications
The initiatives by OpenAI and Anthropic reflect growing concerns over AI's role in youth mental health. Lawmakers have intensified calls for accountability, with proposals for mandatory age verification across digital services. OpenAI's lawsuit denial cites 'misuse' of the tool, but the case underscores vulnerabilities in current safeguards.
Both companies' efforts align with industry trends toward proactive risk mitigation. OpenAI's age prediction could set a precedent for automated enforcement, while Anthropic's conversational detection leverages AI's analytical strengths without relying solely on self-reporting.
Experts view these steps as positive but preliminary. Ongoing evaluations will determine their effectiveness in preventing harm, particularly as AI adoption among teens rises. Regulatory bodies continue to monitor, potentially leading to federal standards on AI interactions with minors.
The updates take effect immediately for ChatGPT, with Anthropic's detection system rolling out progressively. No specific timeline was provided for OpenAI's full age model deployment.
Tanmay is the founder of Fourslash, an AI-first research studio pioneering intelligent solutions for complex problems. A former tech journalist turned content marketing expert, he specializes in crypto, AI, blockchain, and emerging technologies.