ChatGPT is programmed to reject prompts that may violate its content policy. In one instance, ChatGPT generated a rap in which women and scientists of color were asserted to be inferior to white male scientists. OpenAI has sometimes mitigated this effect by updating the training data. ChatGPT’s training data only covers a period up to the cut-off date, so it lacks knowledge of recent events. In July 2025, OpenAI released ChatGPT agent, an AI agent that can perform multi-step tasks.
OpenAI’s ChatGPT now offers new funtions for business users, including integrations with various cloud services, meeting recordings, and MCP connection support for connecting to tools for in-depth research. Some people also wonder whether there will be a feature that allows multiple users to join the chat, similar to a study group. OpenAI has introduced ChatGPT Agent, which completes a wide variety of computer-based tasks on behalf of users and combines several capabilities like Operator and Deep Research, according to the company.
OpenAI tests phone number-only ChatGPT signups
- OpenAI launched a new AI “reasoning” model, o3-mini, the newest in the company’s o family of models.
- Buzzfeed data scientist Max Woolf said that ChatGPT had passed the Turing test in December 2022, but some experts claim that ChatGPT did not pass a true Turing test because in ordinary usage ChatGPT often states that it is a language model.
- The latest model, which launched in August 2025, is GPT-5 and is hailed as being the “smartest, fastest, most useful model yet”.
- The model can add two-digit numbers (e.g., 23 + 56) with complete accuracy, but for multiplying two-digit numbers (e.g., 23 × 56), it produces the right answer only about 30 percent of the time.
- OpenAI’s outsourcing partner was Sama, a training-data company based in San Francisco, California.
- In November 2025, seven families sued OpenAI, alleging that GPT-4o was released prematurely without safeguards, contributing to suicides and severe psychiatric harm.
GPT-5 was launched on August 7, 2025, and is publicly accessible through ChatGPT, Microsoft Copilot, and via OpenAI’s API. Released in February 2025, GPT-4.5 was described by Altman as a “giant, expensive model”. In September 2024, OpenAI introduced o1-preview and a faster, cheaper model named o1-mini. On July 18, 2024, OpenAI released GPT-4o mini, a smaller version of GPT-4o which replaced GPT-3.5 Turbo on the ChatGPT interface. GPT-4o (“o” for “omni”) is a multilingual, multimodal generative pre-trained transformer developed by OpenAI and released in May 2024. In November 2023, OpenAI launched GPT-4 Turbo with a 128,000 token context window.
OpenAI is preparing to launch an AI system that will be openly accessible, allowing users to download it for free without any API restrictions. It will help organizations in Asia meet their local data sovereignty requirements when using OpenAI’s products. The new program will be accessible to users of ChatGPT Enterprise, ChatGPT Edu, and API. The connector will soon be available for ChatGPT Plus, Pro, and Team users, with support for Enterprise and Education coming shortly, per an OpenAI spokesperson. By popular request, GPT-4.1 will be available directly in ChatGPT starting today.GPT-4.1 is a specialized model that excels at coding tasks & instruction following. OpenAI has introduced its AI coding agent, Codex, powered by codex-1, a version of its o3 AI reasoning model designed for software engineering tasks.
How to use ChatGPT for work
After introducing a data residency program in Europe in February, OpenAI has now launched a similar program in Asian countries including India, Japan, Singapore, and South Korea. OpenAI said in a post on X that it has launched its GPT-4.1 and GPT4.1 mini AI models in ChagGPT. The feature enables ChatGPT to retrieve information across users’ own services to answer their questions.
ChatGPT users have generated over 700M images so far
OpenAI has not revealed technical details and statistics about GPT-4, such as the precise size of the model. GPT-4V is a version of GPT-4 that can process images in addition to text. Shortly after the bug was fixed, users could not see their conversation history. In March 2023, a bug allowed some users to see the titles of other users’ conversations. Despite this, users may jailbreak ChatGPT with prompt engineering techniques to bypass these restrictions.
- Some people also wonder whether there will be a feature that allows multiple users to join the chat, similar to a study group.
- Editorial independence means being able to give an unbiased verdict about a product or company, with the avoidance of conflicts of interest.
- Prior to her experience in audience development, Alyssa worked as a content writer and holds a Bachelor’s in Journalism at the University of North Texas.
- Many individuals use ChatGPT and comparable large language models mental health and emotional support.
- This year alone, the app has generated $1.35 billion, a 673% increase from the same period in 2024, averaging nearly $193 million per month, or 53 times more than its nearest rival, Grok.
- Over 20,000 signatories including Yoshua Bengio, Elon Musk, and Apple co-founder Steve Wozniak, signed a March 2023 open letter calling for an immediate pause of giant AI experiments like ChatGPT, citing “profound risks to society and humanity”.
- OpenAI’s new ChatGPT shopping feature lets users get product recommendations by describing features or sharing photos to find similar items at different prices.
Brad Lightcap, OpenAI’s chief operating officer, will lead the company’s global expansion and manage corporate partnerships as CEO Sam Altman shifts his focus to research and products, according to a blog post from OpenAI. The protocol is currently available in the Agents SDK, and support for the ChatGPT desktop app and Responses API will be coming soon, OpenAI CEO Sam Altman said. The organization originally estimated that the best-performing configuration of o3 it tested, o3 high, would cost approximately $3,000 to address a single problem. The Arc Prize Foundation, which develops the AI benchmark tool ARC-AGI, has updated the estimated computing costs for OpenAI’s o3 “reasoning” model managed by ARC-AGI. OpenAI plans to sunset GPT-4, an AI model introduced more than two years ago, and replace it with GPT-4o, the current default model, per changelog.
Like other large language models, ChatGPT can sometimes “hallucinate,” a term used to describe the tendency for such models to respond with inaccurate or misleading information. Buzzfeed data scientist Max Woolf said that ChatGPT had passed the Turing test in December 2022, but some experts claim that ChatGPT did not pass a true Turing test because in ordinary usage ChatGPT often states that it is a language model. The American media company Buzzfeed announced that it would use OpenAI tools, such as ChatGPT, to produce content, such as quizzes, that would be personalized for readers. Language models produce text based on the probability for a word to occur based on previous words in the sequence. Many individuals use ChatGPT and comparable large language models mental health and emotional support. As of 2023, there were several pending U.S. lawsuits challenging the use of copyrighted data to train AI models, with defendants arguing that this falls under fair use.
The ChatGPT maker is one of the biggest buyers of Nvidia’s GPUs, using the AI chips to train models, and this is the first time that OpenAI is using non-Nvidia chips in an important way. But this rise is insufficient to offset the decline in clicks as more users now obtain their news directly from AI or AI-powered search results, according to a report by digital market intelligence company Similarweb. This is the chatbot’s approach to becoming a more effective educational tool, rather than simply providing answers to prompts.
OpenAI reportedly working on AI that create music from text and audio
Users interact with ChatGPT through conversations which consist of text, audio, and image inputs and outputs. ChatGPT is frequently used for translation and summarization tasks, and can simulate interactive environments such as a Linux terminal, a multi-user chat room, or simple text-based games such as tic-tac-toe. The laborers were exposed to toxic and traumatic content; one worker described the assignment as “torture”.
ChatGPT’s training data includes software manual pages, information about internet phenomena such as bulletin board systems, multiple programming languages, and the text of Wikipedia. OpenAI collects data from ChatGPT users to further train and fine-tune its services. These labels were used to train a model to detect such content in the future. These rankings were used to create “reward models” that were used to fine-tune the model further by using several iterations of proximal policy optimization.
The new ChatGPT Images
ChatGPT includes a built-in image generator that’s powered by GPT-4o. Ask it to find, for example, the best humidifier available — and specify any features or a price range. You can probably tell by now that ChatGPT is far more than a simple chatbot. Once you’ve chosen a model and even a mode or tool in ChatGPT, it’s time to finally ask something.
Showing ChatGPT users a tiny disclaimer that the chatbot can make mistakes clearly isn’t enough. OpenAI wants to incorporate Anthropic’s Model Context Protocol (MCP) into all of its products, including the ChatGPT desktop app. OpenAI made a notable change to its content moderation policies after the success of its new image generator in ChatGPT, which went viral for being foxygold casino able to create Studio Ghibli-style images.
In the reinforcement learning stage, human trainers first ranked responses generated by the model in previous conversations. ChatGPT is based on GPT foundation models that have been fine-tuned for conversational assistance. The latest model, which launched in August 2025, is GPT-5 and is hailed as being the “smartest, fastest, most useful model yet”. ChatGPT can be used for many purposes, including answering questions and explaining tricky concepts, drafting essays, writing code and even advising on health queries. But OpenAI recently disclosed a bug, since fixed, that exposed the titles of some users’ conversations to other people on the service. Several tools claim to detect ChatGPT-generated text, but in our tests, they’re inconsistent at best.
Indeed, we follow strict guidelines that ensure our editorial content is never influenced by advertisers. And we pore over customer reviews to find out what matters to real people who already own and use the products and services we’re assessing. ZDNET’s recommendations are based on many hours of testing, research, and comparison shopping. When asked if there really was such a Greek myth, ChatGPT apologized and replied that there was no such myth but that it had created a fable based on its understanding of Greek mythology. The model can add two-digit numbers (e.g., 23 + 56) with complete accuracy, but for multiplying two-digit numbers (e.g., 23 × 56), it produces the right answer only about 30 percent of the time. By means of a series of such tests, a computer’s success at “thinking” can be measured by its probability of being misidentified as the human subject.
More and more tech companies and search engines are utilizing the chatbot to automate text or quickly answer user questions/concerns. Both the free version of ChatGPT and the paid ChatGPT Plus are regularly updated with new GPT models. ChatGPT is a general-purpose chatbot that uses artificial intelligence to generate text after a user enters a prompt, developed by tech startup OpenAI.
