OpenAI CEO Sam Altman has declared a "code red" to focus on improving ChatGPT, according to an internal memo reported by The Wall Street Journal and The Information. The shift delays planned advertising and AI agent products while teams concentrate on the assistant's day-to-day quality. It follows recent Google disclosures that its Gemini 3 models score higher than ChatGPT on several widely cited AI tests.
OpenAI "code red" memo and product changes
According to the reported memo, Altman said ChatGPT's everyday experience needs improvement across personalization, response speed, reliability, and question coverage. OpenAI uses a color-coded priority system, and ChatGPT upgrades have moved from "code orange" to the higher "code red" status. The memo also encourages temporary staff transfers and sets up daily coordination calls for employees driving these changes.
A new reasoning model is expected to launch next week, according to the memo cited by both outlets. OpenAI has not published technical details or a release date for this system on its public channels. The memo frames the new model as part of a broader push to improve ChatGPT responses.
- Advertising tests inside the ChatGPT app are on hold, according to reporting by The Information.
- AI agents for shopping and healthcare use cases are delayed, together with planned updates to the ChatGPT Pulse feature.
- Product and research staff are being redirected toward ChatGPT quality efforts, according to descriptions in the leaked memo.
Competitive context and usage statistics
Google states that its Gemini 3 Deep Think model achieves higher scores than earlier Gemini versions. The company highlights these gains on several reasoning evaluations. One of these is Humanity's Last Exam, a test suite developed by researchers associated with the Center for AI Policy. Google publishes these results on its Gemini 3 Pro benchmark page and related performance documentation.
Independent leaderboards from organizations such as Scale's official leaderboard and Artificial Analysis track Humanity's Last Exam results. These rankings place Gemini 3 Deep Think among the top-performing models on that test. Google's published data shows Gemini 3 Pro scoring higher on this test than several competing systems, including GPT 5. Reporting that Gemini 3 has outscored ChatGPT reflects these documented evaluation results.
Nick Turley, OpenAI's head of ChatGPT, recently said in a public LinkedIn post that ChatGPT is the "number one AI assistant worldwide." He attributed around 70% of assistant usage and roughly 10% of search activity to ChatGPT. The Financial Times has separately reported that OpenAI has more than 800 million weekly users, most on the free tier.
Background and source references
In late 2022 and early 2023, Google leadership declared its own internal "code red" after ChatGPT's rapid adoption. CEO Sundar Pichai redirected teams across Google Research and other units toward accelerated AI product development. The period also saw the launch of Bard in early 2023 and its later evolution into Gemini.
Details of Altman's recent memo come from reports in The Wall Street Journal and The Information. Both publications based their coverage on leaked internal communications. Google's Gemini 3 performance data is drawn from the company's Gemini 3 product blog and Gemini 3 Pro benchmark page. Evaluation scores for Humanity's Last Exam are available on Scale's official leaderboard and on third-party site Artificial Analysis.
Turley's comments about assistant and search usage appear in a recent public LinkedIn post. User estimates for OpenAI and Gemini assistants referenced here follow reporting from the Financial Times. The leaked OpenAI memo also mentions "temporary economic headwinds" and notes Google's recent progress in large-scale AI systems.
.svg)
.svg)
.svg)