• OpenAI CEO Sam Altman has declared a "code red," pausing multiple projects to urgently improve ChatGPT's speed, reliability, and personalization.
  • The move is a direct response to intensifying competition, most notably from Google's Gemini 3, which has reached 650 million monthly users and outperformed GPT-5 on some benchmarks.
  • Internal memos show the company is halting work on ads, shopping tools, health agents, and its Pulse assistant, with teams now joining daily calls to accelerate development.

In a stark internal shift, OpenAI CEO Sam Altman has reportedly issued a "code red" directive to staff, signaling an urgent, all-hands effort to bolster the core capabilities of ChatGPT. According to people familiar with the matter, the move comes as competitive pressure, particularly from Google's surging Gemini platform, intensifies.

The initiative involves a significant reallocation of resources, with several projects being paused to free up engineering talent. Development on advertising integrations, shopping tools, health-focused AI agents, and the company's own Pulse assistant has been temporarily shelved, according to internal communications. The goal is to channel efforts into making ChatGPT faster, more reliable, and significantly more personalized for its user base.

"The focus is now squarely on the core product experience," said one person briefed on the plans. Teams across the organization have been instructed to join daily coordination calls to expedite the rollout of improvements, and employees from paused projects are being encouraged to shift their focus temporarily.

This strategic pivot underscores the rapidly shifting dynamics in the generative AI race. Google's Gemini 3 has not only attracted hundreds of millions of users but has also closed the performance gap, outperforming OpenAI's latest GPT-5 model on certain industry benchmarks. Meanwhile, rivals like Anthropic and DeepSeek continue to apply global pressure, chipping away at OpenAI's once-commanding first-mover advantage.

Despite the competitive headwinds, OpenAI has recently rolled out meaningful upgrades. Its GPT-5 model shows a marked reduction in factual errors, or "hallucinations," and new features like personality customization and improved memory management aim to enhance user stickiness. However, the "code red" suggests leadership believes these incremental gains are insufficient to maintain its edge.

The implications are immediate for OpenAI's workforce and product roadmap. The heightened pace and reprioritization will test organizational agility, while promised features in health and e-commerce will face delays. For the industry, the move highlights that the battleground has expanded beyond raw model capability to encompass user experience, speed, and personalization as critical competitive dimensions.

An OpenAI spokesperson did not immediately respond to a request for comment on the internal directive.

Correction: An earlier version of this article misstated the comparative hallucination rate of GPT-5. It is approximately 45% less likely to contain factual errors than GPT-4o.