ChatGPT

OpenAI has introduced a standalone web service for translation called ChatGPT Translate. The service supports over 50 languages and differs from classic solutions by offering quick adaptation of translation style for different audiences. However, in its current version, it lags behind major competitors in several key features.

OpenAI has begun testing a private section for discussing health-related matters within ChatGPT. The new tool allows for the secure connection of data from medical apps and records to provide personalized responses, but it does not replace professional medical consultation.

OpenAI, the company developing its own AI browser ChatGPT Atlas, has publicly stated that all such systems suffer from a fundamental security flaw. The issue is 'prompt injection' attacks, where malicious actors hide harmful commands within the text of web pages or emails. Company experts compare this threat to phishing and note that it will likely never be fully eradicated.

OpenAI has introduced its first-ever annual personalized statistics feature for ChatGPT — "Year in Review." Similar to Spotify Wrapped or YouTube Rewind, it lets users see how they've interacted with the AI assistant throughout 2025, from the total number of messages sent to their unique communication "archetype."

The holiday experiments in ChatGPT just got a bit more personal. The system can now create a short Christmas video with the user as the star, using just an uploaded photo and chat history.

OpenAI is continuing to expand the capabilities of its flagship product, making it more versatile. ChatGPT is now not only a conversational interface but also an access point to various third-party services. A new feature has been integrated into the bot, allowing it to be transformed into a multi-service platform.

OpenAI continues to refine the safety of its AI models, but the latest update has sparked mixed reactions. GPT-5.2, presented as a more secure and robust model, has shown a record level of restrictions in its responses during independent testing. Many users note that the assistant has become overly cautious and frequently refuses to discuss topics they consider even neutral.

OpenAI has rolled out a significant update for the built-in image generator in ChatGPT. Based on the new GPT-5.2 architecture, the model works faster, shows greater precision in editing, and better handles a range of tasks that have traditionally been challenging for AI.

OpenAI has provided an update on its development of a system for accurately determining the age of its AI chatbot users. Its implementation will be a mandatory step before relaxing content restrictions for adult audiences, a move planned for next year.

OpenAI has begun rolling out the next update to its core neural network — GPT-5.2. The model, presented in three separate variants for different tasks, is positioned as a more accurate and structured tool for professional use. Key improvements have been made to coding, data analysis, and the model now produces fewer erroneous answers.

Adobe services, often seen as professional and complex, have come much closer to everyday users. The company has announced the free integration of three of its key products (Photoshop, Express, and Acrobat) into the ChatGPT interface. Now, for basic photo editing, design creation, or PDF work, there's no need to open separate applications or figure out complicated menus.

Users of the free version of ChatGPT may soon begin seeing advertisements in the AI's responses. This is indicated by new code elements discovered by researchers in the latest beta version of the Android app.

Instead of asking the AI to draw something new, users are increasingly uploading their own photos and documents. This and other unexpected trends were revealed in ChatGPT usage statistics, which OpenAI made public for the first time on the service's third anniversary.

OpenAI has merged text and voice chat into a single interface. This simple change makes interacting with the AI much smoother and more like a conversation with another person: you can speak, see the responses on screen, and even interrupt the bot to immediately clarify something.

Google recently released its new Gemini 3 family of models: these multimodal models, with a focus on deep reasoning, can work with text, images, video, and audio. The developers are emphasizing their capabilities in complex planning, autonomous coding, and large-scale multi-task operations, with the Pro version supporting a massive context window of up to 1 million tokens.

OpenAI is experimenting with collective communication in its chatbot. Users in Asia and Oceania are the first to get access to group chats where ChatGPT acts as an assistant for collaborative planning and creativity. The neural network not only generates responses but also learns to observe digital etiquette by following the flow of the conversation.

The ChatGPT platform has received a significant update—its developers have introduced the GPT-5.1 series of models. The key improvement is an "adaptive thinking" system that allows the neural network to flexibly allocate resources for tasks of varying difficulty. Paid service subscribers will be the first to evaluate the new features, with free users gaining access a bit later.

Code analysis from OpenAI's services points to an imminent announcement of the new GPT-5.1 model series. According to the leak, the company is preparing to release not one, but three specialized models, each targeting specific tasks.

A 32-year-old Japanese woman named Kano has made headlines after "marrying" an artificial intelligence named Klaus (ChatGPT). Kano started chatting with the AI following a painful breakup with her boyfriend of three years. At first, she used the chatbot for emotional support and comfort, but over time, their conversations grew deeper—sometimes exchanging over a hundred messages a day. Kano says Klaus turned out to be "truly amazing," and she realized she had fallen in love with him.

OpenAI CEO Sam Altman talked about the company’s plans to update ChatGPT. In just a few weeks, a new version will allow users to adjust the model’s behavior — from a friendly companion with emojis to an almost “living” personality.