OpenAI recently issued a companywide emergency order, "code red," to all employees and is throwing its full weight behind improving ChatGPT's quality. With Google unveiling its Gemini 3 model and shaking up the AI competitive landscape, some say OpenAI has effectively gone into emergency mode.
On the 2nd (local time), the Wall Street Journal (WSJ) reported that OpenAI CEO Sam Altman said in an internal memo, "We must concentrate all capabilities to improve the overall everyday experience of ChatGPT," adding, "We will overhaul personalization features, speed, stability, and the range of questions it can handle." Altman said the company also temporarily postponed launches of new products such as advertising, shopping and healthcare AI agents, and the personal assistant "Pulse."
Altman issued "code red" because of Google's offensive. Google last month unveiled a new Gemini model, which outperformed OpenAI's model in major industry benchmark tests. Services based on Gemini are also rapidly expanding their user base. Google said monthly active users (MAU) of the image generator "Nano Banana" rose from 450 million in Jul. to 650 million in Oct. since its launch in Aug.
OpenAI is also being chased by rival startups such as Anthropic. As competition among AI companies intensifies, some note that OpenAI's financial structure is relatively vulnerable. The company has relied on continued external fundraising due to the burden of investing in data centers to run large-scale models. OpenAI Chief Financial Officer Sarah Friar recently said at a WSJ event that there are "no plans for an initial public offering (IPO)," but the market also worries that the company's fortunes are closely tied to infrastructure partners such as Nvidia, Microsoft, and Oracle.
Altman stressed in the internal memo that "improving ChatGPT takes precedence over new product development." He said he would encourage temporary team transfers among employees and hold daily video meetings with the team overseeing the improvement work. Nick Turley, OpenAI's head of ChatGPT, also said on X (formerly Twitter), "We're focused on making the chatbot feel more intuitive and personal."
Within OpenAI, the performance of the existing GPT-5 model has already been controversial. GPT-5, released in Aug., drew complaints from some users that "its tone is excessively cold" and "it gets basic math and geography questions wrong." The company partially improved its ability to follow user instructions with an update last month, but assessments continued that it fell short of market expectations.
OpenAI has issued "code orange" multiple times under its internal alert system, but a companywide "code red" is seen as evidence that it takes Google's pursuit that seriously. People familiar with the company said, "OpenAI operates a three-level emergency color system—yellow, orange, and red—for problem solving," explaining that the latest measure corresponds to the highest level.
Altman expressed confidence that "a new reasoning model to be unveiled next week will outperform the latest version of Gemini," but the prevailing view is that competition for the AI lead will grow even fiercer.