OpenAI Pushes New ChatGPT Safety Features as Lawsuits Mount

OpenAI connected Thursday announced caller information features designed to assistance ChatGPT admit signs of escalating hazard crossed conversations arsenic the institution faces increasing ineligible and governmental scrutiny implicit however its chatbot handles users successful distress.

In a blog post, OpenAI said the updates amended ChatGPT’s quality to place informing signs tied to suicide, self-harm, and imaginable unit by analyzing discourse that develops implicit clip alternatively of treating each connection separately.

“People travel to ChatGPT each time to speech astir what matters to them—from mundane questions to much idiosyncratic oregon analyzable conversations,” the institution wrote. “Across hundreds of millions of interactions, immoderate of these conversations see radical who are struggling oregon experiencing distress.”

According to OpenAI, ChatGPT present uses impermanent “safety summaries,” which it described arsenic narrowly scoped notes that seizure applicable safety-related discourse from earlier conversations.

“In delicate conversations, discourse tin substance arsenic overmuch arsenic a azygous message,” the institution wrote. “A petition that appears mean oregon ambiguous connected its ain whitethorn transportation a precise antithetic meaning erstwhile viewed alongside earlier signs of distress oregon imaginable harmful intent.”

OpenAI said the summaries are short-term notes utilized lone successful superior situations, not to permanently retrieve users oregon personalize chats, and are utilized to spot signs that a speech is becoming dangerous, debar giving harmful information, de-escalate the situation, oregon usher users toward help.

“We focused this enactment connected acute scenarios, including suicide, self-harm, and harm to others,” they wrote. “Working with intelligence wellness experts, we updated our exemplary policies and grooming to amended ChatGPT’s quality to admit informing signs that look implicit the people of a speech and usage that discourse to pass much cautious responses.”

The announcement comes arsenic OpenAI faces aggregate lawsuits and investigations alleging ChatGPT failed to decently respond to unsafe conversations involving violence, affectional vulnerability, and risky behavior.

In April, Florida Attorney General James Uthmeier launched an investigation into OpenAI tied to concerns astir kid safety, self-harm, and the 2025 wide shooting astatine Florida State University. OpenAI is besides facing a national suit alleging ChatGPT helped the suspected gunman transportation retired the attack.

On Tuesday, OpenAI and CEO Sam Altman were sued successful California authorities tribunal by the household of a 19-year-old pupil who died from an accidental overdose, with the suit alleging ChatGPT encouraged unsafe cause usage and advised connected mixing substances.

OpenAI said helping ChatGPT admit “risk that lone becomes wide implicit time” remains an ongoing challenge; akin information methods could yet grow into different areas.

“Today, this enactment focuses connected self-harm and harm-to-others scenarios. In the future, we whitethorn research whether akin methods tin assistance successful different high-risk areas specified arsenic biology oregon cyber safety, with cautious safeguards successful place,” they wrote. “This remains an ongoing priority, and we volition proceed strengthening safeguards arsenic our models and knowing evolve.”