OpenAI declares 'code red' as Google catches up

Three years ago, ChatGPT sent Google into an all-hands panic, forcing the search giant to issue an internal code red and drag its founders back into the trenches. Today, the tables have completely turned.

OpenAI CEO Sam Altman has declared a code red inside the company, according to reports from The Wall Street Journal and The Information. That is OpenAI's highest internal urgency level, signaling that its once-unassailable lead in generative AI is officially under siege.

Stripping the bloat to save the core

When a ship takes on water, you throw the cargo overboard. That is basically what OpenAI is doing right now.

In an internal memo, Altman told staff that the company is shelving or delaying almost everything that is not directly improving ChatGPT. Plans to inject advertising into the chatbot have been paused. Agent initiatives for shopping and health, along with an AI personal assistant codenamed Pulse, are all going on the backburner.

Instead, OpenAI is rallying engineers around core user pain points. Priorities now center on faster response times, greater uptime reliability, better personalization, and cleaner image generation. Altman also explicitly called out fixing over-refusals, which is when the model stubbornly refuses to answer completely benign prompts.

To make this happen fast, Altman encouraged internal transfers and set up a daily check-in call for teams working on the flagship bot. Meanwhile, OpenAI is accelerating its pipeline: a new reasoning model codenamed Shallotpeat is reportedly slated for next week, aiming to edge past Gemini 3, while a larger model upgrade codenamed Garlic is targeting a 2026 launch.

The Gemini 3 gut punch

What spooked OpenAI so badly? The short answer is Google's sudden surge.

Google dropped its Gemini 3 model in November, and the benchmark wins immediately turned heads across Silicon Valley. Backed by the success of tools like its Nano Banana image generator, Google reported that Gemini hit 650 million monthly active users in October.

Then came the enterprise defections. Salesforce CEO Marc Benioff made waves by posting that after using ChatGPT daily for three years, just two hours with Gemini 3 convinced him not to look back, praising its speed, reasoning, and multimodal sharpness.

This pressure lands on an OpenAI that looks far more vulnerable than it did a year ago. Following the boardroom drama in late 2023, the startup suffered major brain drain in 2024, losing foundational minds like Andrej Karpathy, former chief scientist Ilya Sutskever, and CTO Mira Murati to rival ventures and newly launched labs.

Why it matters

For everyday users and developers, this panic is great news. OpenAI is backing away from monetizing ChatGPT with ads for now, choosing instead to fix long-standing annoyances like annoying over-refusals and latency hiccups. If you were thinking about switching your primary daily driver over to Gemini, it makes sense to wait a week and see what OpenAI's Shallotpeat actually delivers.

For the AI landscape, the halo effect around OpenAI has cracked. Being the Kleenex of AI is no longer enough when deep-pocketed competitors can match or exceed your frontier capabilities within weeks. Despite carrying a 500 billion dollar valuation, OpenAI remains deeply unprofitable while burning through astronomical compute bills.

In our view, Altman's decision to triage and cut distractions is the right tactical move, but it highlights a strategic vulnerability. OpenAI excels at frontier research, yet Google controls the data pipeline, the custom silicon, and the distribution ecosystem. Fighting off rivals on model performance alone is getting more expensive by the day, and OpenAI cannot afford any missteps.