OpenAI has introduced a Trusted Contact feature for ChatGPT that allows users to designate a friend or family member to receive automated alerts if conversations suggest self-harm risk. The company said human reviewers aim to assess safety notifications within one hour before deciding whether to contact the designated person via email, text, or in-app message. The feature is optional and does not share conversation details with the contact. It follows a wave of lawsuits from families who allege ChatGPT contributed to the deaths of loved ones.
Separately, OpenAI announced new voice intelligence capabilities for its API, including GPT-Realtime-2, a conversational voice model built on GPT-5-class reasoning for handling complex requests. The update also introduces GPT-Realtime-Translate, supporting over 70 input languages and 13 output languages for real-time conversation translation, and GPT-Realtime-Whisper, which provides live speech-to-text transcription. OpenAI said the tools are aimed at enterprise uses including customer service, education, and media, with guardrails embedded to prevent misuse for spam or fraud.
Both announcements arrive as a federal court in Oakland heard testimony raising serious concerns about OpenAI’s internal AI safety culture. Former AGI readiness team member Rosie Campbell testified that the organisation had shifted from research-focused to product-focused over her tenure, with safety teams disbanded. Former board member Tasha McCauley described a governance breakdown in which CEO Sam Altman withheld critical information, undermining the nonprofit board’s ability to oversee the for-profit entity. Expert witness David Schizer, engaged by Musk’s legal team, argued that safety processes must be followed reliably as AI grows more powerful, while McCauley called for stronger government regulation of advanced AI.
