OpenAI to Enhance ChatGPT Safety Features -- THE Journal

Artificial Intelligence

OpenAI to Enhance ChatGPT Safety Features

By John K. Waters
09/03/25

OpenAI has announced it is strengthening safety measures within ChatGPT to better detect and respond to users experiencing mental health crises. The artificial intelligence company said it will update ChatGPT to recognize various forms of mental distress and improve safeguards around mental health-related conversations, which can deteriorate during prolonged chat sessions. The changes include better detection of concerning behavior, such as identifying when users express feelings of invincibility after sleep deprivation.

Technical Challenges

OpenAI acknowledged that its current safeguards work effectively in short conversations but can become less reliable during extended interactions. The company stated that its safety protocols may become less effective as conversations lengthen, potentially allowing harmful content to slip through that would normally be blocked.

The company is developing improvements to maintain safety measures across long conversations and multiple chat sessions. ChatGPT's ability to reference previous conversations presents additional challenges for maintaining consistent safety protocols.

Planned Improvements

OpenAI outlined several planned enhancements:

Mental Health Response: The company will expand interventions beyond acute self-harm cases to address other forms of mental distress. Updates will train ChatGPT to de-escalate concerning situations by grounding users in reality.
Emergency Services Access: The company plans to provide one-click access to emergency services and is exploring connections to certified therapists and licensed professionals through the platform.
Parental Controls: New features will enable parents to monitor and control their teenage children's use of ChatGPT, including options for emergency contact designation.
Global Resource Expansion: OpenAI is localizing mental health resources beyond the U.S. and Europe to serve international users.

Current Safety Measures

OpenAI noted ChatGPT currently includes several safety features:

Training to recognize self-harm expressions and respond with empathetic language while directing users to professional help;
Automatic blocking of responses that violate safety guidelines, with stronger protections for minors;
Referrals to suicide prevention hotlines: 988 in the U.S., Samaritans in the U.K., and findahelpline.com elsewhere; and
Human review of cases involving potential harm to others.

The company works with more than 90 physicians across 30 countries and maintains an advisory group of mental health experts, youth development specialists, and human-computer interaction researchers.

Market Impact

ChatGPT, launched in late 2022, catalyzed the current generative AI boom and maintains more than 700 million weekly users. The platform has expanded beyond initial use cases to include personal advice, coaching, and emotional support conversations.

OpenAI recently deployed GPT-5 as ChatGPT's default model, claiming more than 25% reduction in problematic responses during mental health emergencies compared to its predecessor.

OpenAI said it had planned to detail its mental health response improvements after its next major update but decided to share information earlier due to "recent heartbreaking cases of people using ChatGPT in the midst of acute crises."

About the Author

John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].

E-Mail this page

Printable Format

Featured

Establishing a Proactive Defense Against Evolving Cyber Threats

Here are six good starting points for K-12 districts that want to improve their cybersecurity mitigation strategies and take a more proactive approach to mitigating risk.
Survey: AI Optimism Is Rising, but Cheating and Privacy Concerns Persist

Artificial intelligence is evolving rapidly — both in how it's used and how it's perceived in K-12 education. As a result, schools and districts are under increasing pressure to adapt and respond to the changes AI is driving.
You've Paid for Tutoring. Here's How to Make Sure It Works.

As districts and states nationwide invest in tutoring, it remains one of the best tools in our educational toolkit, yielding positive impacts on student learning at scale. But to maximize return on investment, both financially and academically, we must focus on improving implementation.
Boise School District Boosts Information Literacy with Gale In Context Databases

Students in Idaho's Boise School District are learning how to analyze and think critically about the information they read through the use of Gale In Context.