According to an internal memo, OpenAI CEO Sam Altman has announced to all employees that the company is immediately entering a "red alert" state and will allocate more resources to fully optimize ChatGPT in order to cope with the increasing competitive pressure from Google and other AI competitors.
Altman stated that, as a result, the company will postpone a series of non-core businesses, including advertising.
“ChatGPT is at a critical turning point,” he emphasized.
Although OpenAI has never publicly acknowledged that it is developing an advertising business, sources familiar with the matter have revealed that the company is testing various advertising formats, including online shopping ads. Currently, millions of users have searched for and purchased goods through ChatGPT.
Altman pointed out that launching this "red alert" special task force means that OpenAI will temporarily suspend the development of several product lines, including AI agents designed to automate shopping and health tasks, and the "Pulse" function that can generate personalized morning briefings for ChatGPT users.
The memo reveals that OpenAI previously declared an "orange alert" state to improve ChatGPT. According to sources, the company uses three different color codes (yellow, orange, and red) to describe varying degrees of urgency in resolving issues.
Altman did not specify the current problems facing ChatGPT, but Google disclosed this fall that its chatbot Gemini's user base is continuing to expand. It is understood that Altman recently privately warned employees that Google's strong comeback in the AI field may put OpenAI under "temporary economic pressure."
During a conference call with OpenAI investors last month, CFO Sarah Friyal mentioned the slowdown in ChatGPT's growth but did not provide specifics.
ChatGPT's performance is directly related to OpenAI's ability to successfully raise approximately $100 billion to support the company's anticipated massive operational investments. This summer, OpenAI predicted that despite needing to invest tens of billions of dollars in new technology research and development and the operation of products like ChatGPT over the next few years, the chatbot is expected to generate approximately $10 billion in revenue this year through subscription services, reaching $20 billion next year, and potentially increasing to $35 billion by 2027.
OpenAI's recent activation of the "red alert" is essentially a role reversal. Three years ago, Google activated a "red alert" due to the threat posed to its search engine by ChatGPT. Subsequently, Google launched the Gemini chatbot, which, although still lagging behind OpenAI in user scale, has shown signs of catching up.
Data released by Google in October showed that, although there is still a significant gap compared to the number of users of ChatGPT announced by OpenAI, the number of monthly active users of Gemini has grown from 450 million in July to 650 million.
In addition, Google has launched an "AI mode" in its search engine, which essentially transforms the search application into an interactive chat interface similar to ChatGPT.
Regarding OpenAI, Nick Terry, head of ChatGPT, posted that according to the company's estimates, ChatGPT currently handles about 70% of AI "assistant" interactions and 10% of "search" requests globally.
Altman also stated in the memo that he has reassigned more staff to focus on improving the functionality and experience of ChatGPT, with one of the key focuses being providing deep personalization support for 800 million weekly active users, including allowing users to customize their interaction methods.
In addition, key tasks during the "red alert" phase include enhancing the Imagegen image generation model, which allows users to create a variety of visual content, from interior design sketches to turning live-action photos into animated works. Notably, Google launched its self-developed image generation model, Nano Banana Pro, last month, which has received positive market feedback.
Ultraman also listed other key areas for optimization:
- Improve model behavior to enhance user preference for the underlying ChatGPT model, including its performance on public evaluation leaderboards such as LMARaena;
- Improve the response speed and stability of ChatGPT;
- Efforts should be made to reduce the phenomenon of "excessive rejection," that is, to prevent chatbots from refusing to answer reasonable questions without cause.
Ultraman issued a "red alert" just as new models released by competitors such as Google and Anthropic were receiving high praise from the developer community.
Altman revealed that OpenAI is accelerating the development of a new generation of large language models called Garlic in order to cope with the recent competitive pressure of being considered to be lagging behind Google in the field of AI, and to strive to regain its technological leadership.
According to sources familiar with the matter, OpenAI's Chief Research Officer, Mark Chen, stated in an internal communication last week that Garlic performed well in the company's internal evaluations, surpassing Google's Gemini 3 and Anthropic's Opus 4.5, at least in key tasks such as programming and inference. The company plans to release its version as soon as possible, meaning GPT-5.2 or GPT-5.5 may be released in early 2026.
Garlic is a new model independent of Shallotpeat. Shallotpeat is a new large language model that OpenAI is developing, and Altman told employees in October that Shallotpeat would help OpenAI challenge Gemini 3. Garlic includes bug fixes that the company used during the pre-training process of developing Shallotpeat.
Altman acknowledged that the company still has a lot of work to do to improve the overall user experience of ChatGPT. These reasoning models, by utilizing stronger computing power to generate higher-quality answers, provide technical support for ChatGPT's "thinking modes" and "deep research" functions.
This article is from Tencent Technology , authored by Jin Lu, and published with authorization from 36Kr.
