A new beta build of the ChatGPT Android app includes references to ad features, suggesting that OpenAI is working on an advertising layer for its chatbot. The code strings appear soon after reports that the company is exploring ad formats that may use ChatGPT’s memory and past chats for targeting, despite earlier public comments from CEO Sam Altman that described ads as a last resort.
A new study from Italy’s Icaro Lab finds that simply rewriting dangerous requests as poems can bypass safety filters in many leading AI chatbots. The researchers report a 62 percent success rate in getting prohibited responses from 25 large language models, raising fresh concerns about how easily creative prompt attacks can undermine AI safety systems.
OpenAI has confirmed a data breach at analytics partner Mixpanel exposed limited developer information linked to its API platform, while stressing that ChatGPT users and core OpenAI systems were not affected.
Offering ChatGPT for free to teachers is not a goodwill move aimed at education alone. It is a structural attempt to shape how AI enters classrooms, how teachers retain authority over its use, and how schools avoid fragmented, unsupervised adoption. The decision reflects constraints around trust, misuse, training load, and institutional inertia rather than a simple push for scale.
Target is launching a new ChatGPT integration that lets shoppers search, build carts, and check out using their Target account directly within the chatbot, with support for shipping, pickup, and upcoming same-day delivery.
OpenAI has introduced GPT 5.1 as an updated default model family in ChatGPT, combining GPT 5.1 Instant and GPT 5.1 Thinking and using adaptive reasoning to select response depth. The rollout starts with paid tiers and expands to other users, while new personality controls allow tone changes during an active conversation.
OpenAI has added a feature that allows users to interrupt ChatGPT mid-response and adjust the prompt without restarting the entire conversation. The change offers a more practical way to guide long, detailed answers and reduces wasted time when the model takes several minutes to respond.
YouTube is introducing a likeness detection tool to help identify and remove AI-generated deepfakes that misuse people’s faces, marking an early step toward better control over digital identity.











