In November 2023, OpenAI launched GPT-4 Turbo with a 128,000 token context window. A study presented example attacks on ChatGPT, including jailbreaks and reverse psychology. ChatGPT is programmed to reject prompts that may violate its content policy. The term "hallucination" as applied to LLMs is distinct from its meaning in psychology, and the phenomenon in chatbots is more similar to confabulation or bullshitting.
GPT-4
Despite this, users may "jailbreak" ChatGPT with prompt engineering techniques to bypass these restrictions. … It's also a way to understand the "hallucinations", or nonsensical answers to factual questions, to which large language models such as ChatGPT are all too prone. In one instance, ChatGPT generated a rap in which women and scientists of color were asserted to be inferior to white male scientists. The reward model of ChatGPT, designed around human oversight, can be over-optimized and thus hinder performance, in an example of an optimization pathology known as Goodhart's law.
- OpenAI acknowledged that there have been "instances where our 4o model fell short in recognizing signs of delusion or emotional dependency", and reported that it is working to improve safety.
- ChatGPT has never been publicly available in China because OpenAI prevented Chinese users from accessing their site.
- In March 2023, ChatGPT Plus users got access to third-party plugins and a browsing mode (with Internet access).
- In December 2022, the question-and-answer website Stack Overflow banned the use of ChatGPT for generating answers to questions, citing the factually ambiguous nature of its responses.
- Google’s Bard (now Gemini) launched on February 6, 2023, one day before Microsoft’s announcement of Bing Chat (now Microsoft Copilot).
- One study analyzed ChatGPT’s responses to 517 questions about software engineering or computer programming posed on Stack Overflow for correctness, consistency, comprehensiveness, and concision.
Download the YouTube mobile app
ChatGPT's Mandarin Chinese abilities were lauded, but the ability of the AI to produce content in Mandarin Chinese in a Taiwanese accent was found to be "less than ideal" due to differences between mainland Mandarin Chinese and Taiwanese Mandarin. None of the tested services were a perfect replacement for a fluent human translator. In December 2023, the Albanian government decided to use ChatGPT for the rapid translation of European Union documents and the analysis of required changes needed for Albania's accession to the EU. ChatGPT (based on GPT-4) was better able to translate Japanese to English when compared to Bing, Bard, and DeepL Translator in 2023.
GPT-4.5
TrendForce market intelligence estimated that 30,000 Nvidia GPUs (each costing approximately $10,000–15,000) were used to power ChatGPT in 2023. On August 19, 2025, OpenAI launched ChatGPT Go in India, a low-cost subscription plan priced at ₹399 per month, offering ten times higher message, image generation, and file-upload limits, double the memory span compared to the free version, and support for UPI payments. In October 2023, OpenAI's image generation model DALL-E 3 was integrated into ChatGPT Plus and ChatGPT Enterprise. According to the company, the paid version of the website was still experimental, but provided access during peak periods, no downtime, priority access to new features, and faster response speeds.
App support
In August 2024, OpenAI announced it had created a text watermarking method but did not release it for public use, saying that users would go to a competitor without watermarking if it publicly released its watermarking tool. OpenAI CEO Sam Altman said that users were unable to see the contents of the conversations. In October 2025, OpenAI reported that approximately 0.15% of ChatGPT's active users in a given week have conversations including explicit indicators of potential suicidal planning or intent, translating to more than a million people weekly.
In February 2023, OpenAI launched a premium service, ChatGPT Plus, that costs US$20 per month. ChatGPT was initially free to the public, and OpenAI planned to monetize the service later. In October 2025, OpenAI launched ChatGPT Atlas, a browser integrating the ChatGPT assistant directly into web navigation, to compete with existing browsers such as Google Chrome and Apple's Safari.
In March 2023, ChatGPT Plus users got access to third-party plugins and a browsing mode (with Internet access). Premium users were originally limited in the number of messages they could send to the new model, but OpenAI increased and eventually removed these limits. The model can also generate new images based on existing ones provided in the prompt, which can, for example, be used to transform images with specific styles or inpaint areas. In December 2024, OpenAI launched a new feature allowing users to call ChatGPT with a telephone for up to 15 minutes per month for free. OpenAI collects data from ChatGPT users to further train and fine-tune its services.
OpenAI said it has taken steps to effectively clarify and address the issues raised; an age verification tool was implemented to ensure users are at least 13 years old. Italian regulators assert that ChatGPT was exposing minors to age-inappropriate content, and that OpenAI's use of ChatGPT conversations as training data could violate Europe's General Data Protection Regulation. ChatGPT also provided an outline of how human reviewers are trained to reduce inappropriate content and to attempt to provide political information without affiliating with any political position. ChatGPT gained one million users in five days and 100 million in two months, becoming the fastest-growing internet application in history. In The Atlantic magazine's "Breakthroughs of the Year" for 2022, Derek Thompson included ChatGPT as part of "the generative-AI eruption" that "may change our mind about how we work, how we think, and what human creativity is".
These labels were used to train a model to detect such content in the future. To build a safety system against harmful content (e.g., sexual abuse, violence, racism, sexism), OpenAI used outsourced Kenyan workers earning around $1.32 to $2 per hour to label such content. 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 ethics of its development, cycling classes particularly the use of copyrighted content as training data, have also drawn controversy.
Between March and April 2023, Il Foglio published one ChatGPT-generated article a day on its website, hosting a special contest for its readers in the process. The chatbot technology can improve security by cyber defense automation, threat intelligence, attack identification, and reporting. In July 2024, Futurism reported that GPT-4o in ChatGPT would sometimes link "scam news sites that deluge the user with fake software updates and virus warnings"; these pop-ups can be used to coerce users into downloading malware or potentially unwanted programs. Chris Granatino, a librarian at Seattle University, noted that while ChatGPT can generate content that seemingly includes legitimate citations, in most cases those citations are not real or largely incorrect.