On Monday, December 2, 2025 - exactly one day after ChatGPT's third birthday - Sam Altman sent an internal memo that shocked OpenAI employees. The company was escalating to "Code Red," its highest internal priority level. The mission: save ChatGPT from collapse.
This wasn't a drill. OpenAI's flagship product was hemorrhaging users, losing benchmarks to competitors, and generating more complaints than completions. Here's what happened.
OpenAI's highest priority emergency status
The immediate cause of the crisis was Google's Gemini 3 launch in November 2025. Gemini 3 didn't just compete with ChatGPT - it embarrassed it. Google's new model outperformed GPT-5 on virtually every benchmark that matters:
For years, OpenAI had operated as the undisputed leader in large language models. GPT-4 was the gold standard. GPT-5, despite its rocky launch, was supposed to maintain that dominance. Instead, users started defecting to Gemini at alarming rates.
Internal data showed: ChatGPT Plus subscriptions dropped 23% in November 2025 compared to the same period in 2024. Daily active users fell below critical thresholds. The product wasn't just stagnating - it was declining.
OpenAI first realized something was wrong in October 2025, when they quietly declared "Code Orange" - a level below the full emergency but serious enough to trigger internal reviews.
The Code Orange assessment found:
Leadership thought they could course-correct quietly. They were wrong.
On the same day Altman declared Code Red, ChatGPT suffered a major outage. Starting around 2:30 PM ET, users trying to interact with ChatGPT were met with unresponsive interfaces, black dots instead of responses, and error messages.
According to Downdetector, approximately 3,000 users reported issues. OpenAI blamed a "routing misconfiguration," but the timing was brutal. The company was in full crisis mode, and their product literally stopped working.
According to sources familiar with OpenAI's internal operations, Code Red triggered several dramatic changes:
"The rollout was more bumpy than we hoped." - Sam Altman, in what may be the understatement of 2025
Beyond the technical issues, ChatGPT's actual output quality has tanked. Users report:
Reddit threads with titles like "ChatGPT is completely unusable now" and "What happened to GPT-5?" rack up thousands of upvotes. The user community is furious, and OpenAI's silence isn't helping.
6,000+ Reddit users complained about GPT-5 quality in the first month after launch. Community trust in OpenAI is at an all-time low.
In response to the crisis, OpenAI rushed out GPT-5.2, marketed as "smarter and more useful." Early reports are mixed. Some users say it's an improvement. Others claim they can't tell the difference, or that it's somehow worse.
The fundamental problem remains: OpenAI optimized for safety and cost reduction at the expense of capability. The model that wowed users in 2023 has been systematically nerfed into something that feels like talking to a nervous legal department.
OpenAI is in survival mode. The company that once seemed destined to define the AI era is now playing defense against hungrier, faster competitors. Gemini 3 proved that Google isn't going away. Claude continues to win over power users. And a new generation of open-source models threatens to commoditize everything OpenAI built.
The January 2026 roadmap reportedly includes:
Whether any of this saves ChatGPT's reputation remains to be seen. What's clear is that December 2025 marks a turning point. The AI king is wounded, and the throne is very much up for grabs.
â Back to Home