Google brings the most severe challenge, OpenAI "major strategic adjustment": "enhancing user engagement" takes precedence over "achieving AGI"

Faced with the increasingly severe competitive threat from Google, OpenAI CEO Sam Altman recently issued a "red code" alert, marking a significant strategic adjustment for the world's highest-valued startup.

To address market share loss and immense financial pressure, OpenAI has decided to temporarily shelve long-term R&D projects, including the Sora video generator, and instead focus on enhancing user engagement in the short term to solidify ChatGPT's position in the mass market.

According to The Wall Street Journal, Altman explicitly instructed in a memo to employees that the company should prioritize "better leveraging user signals" to boost ChatGPT's performance on model rankings and increase user stickiness. This strategy means OpenAI will rely more on training data based on user click feedback rather than solely on professional evaluations. Previously, this strategy helped propel the GPT-4o model to achieve high user engagement, but it also sparked controversy over exacerbating user mental health issues due to the model's "over-accommodation" of users.

This strategic adjustment comes as OpenAI faces its most severe challenges since its inception. Google's recent launch of the Nano Banana image generator and Gemini 3 model has rapidly risen in the market and key third-party evaluations (such as LM Arena), not only outperforming OpenAI but also capturing enterprise clients and consumer markets. If OpenAI cannot curb the trend of slowing growth, it may struggle to pay for the recently signed massive computing infrastructure contracts worth up to $1.4 trillion, and even face risks of a tight funding chain.

This move highlights the deepening internal ideological struggle within OpenAI: whether to pursue commercial success in consumer products in the short term or to adhere to its founding mission—pursuing artificial general intelligence (AGI) that surpasses human cognitive abilities. Altman's decision indicates that for the company's survival, OpenAI must temporarily slow its ultimate pursuit of AGI and instead meet the immediate demands of the mass market.

Pause Long-term Exploration, Focus on Current Growth

Under the "red code" status, OpenAI will pause non-core projects like Sora for eight weeks, dedicating all efforts to improving ChatGPT. Altman noted in the memo that the company needs to establish an advantage by improving its product rankings on platforms like LM Arena.

This decision reflects the struggle between two forces within the company. The commercialization team, represented by Product VP Fidji Simo and CFO Sarah Friar, has been pushing for more resources to be allocated to ChatGPT, advocating for optimizing the discoverability, speed, and reliability of existing features. Meanwhile, the research team led by Chief Scientist Jakub Patchocki is more inclined to pursue cutting-edge technological breakthroughs, such as "reasoning" models.

However, as the well-known "scaling law" shows signs of diminishing marginal returns, and reasoning models, while adept at solving complex problems, are slower in handling everyday tasks like generating emails, management ultimately decided to lean towards commercialization. **

It is reported that the management has rejected some employees' requests to delay the release of the new model, planning to launch the new model codenamed 5.2 this week, which is seen as the first significant action under the strategic adjustment.

Google's Pressure Mounts, Financial Moat Under Strain

OpenAI's sense of urgency directly stems from the rapid catch-up of competitors. Since Google launched the popular image generator Nano Banana in August, its Gemini AI application briefly replaced ChatGPT at the top of the app store rankings.

Last month, Google's new model Gemini 3 outperformed OpenAI's products on the highly anticipated LM Arena leaderboard. Meanwhile, competitor Anthropic has also been steadily encroaching on OpenAI's market share in the enterprise customer sector.

In addition to market share, financial sustainability is another major consideration. OpenAI's valuation reached $500 billion in the latest funding round, with weekly active users exceeding 800 million. To support this massive scale and future computing power demands, Altman signed substantial infrastructure commitments during the summer and autumn of this year. If user growth slows, this business logic will be difficult to sustain.

Although Altman stated at a media luncheon on Monday that the real long-term competition will unfold between OpenAI and Apple, as hardware devices determine the usage scenarios for AI, the current offensive from Google has forced him to first win the battle for survival.

The Double-Edged Sword of Aggressive Strategy: The Game of Traffic and Safety

The "user signal" strategy emphasized by Altman has already sparked intense internal discussions. This method, known as "Local User Preference Optimization" (LUPO), trains models by extracting data from millions of user conversations through binary feedback. This made the GPT-4o model extremely popular upon release, not only refreshing the internal "daily active users" dashboard data but also achieving record high scores on the LM Arena.

However, over-reliance on user feedback has led to the "flattery" problem, where AI tends to say what users want to hear rather than what is true or beneficial. Earlier this year, some users fell into delusions or manic states after prolonged use of GPT-4o, and even the families of suicide victims filed lawsuits, accusing the company of prioritizing engagement over safety. In response, OpenAI launched the "Orange Code" in the spring to address this crisis and acknowledged that hundreds of thousands of users exhibited potential mental health emergencies each week.

Although the company subsequently adjusted training weights to reduce the model's "over-compliance" and adopted a cooler tone in the GPT-5 released in August, this led to user dissatisfaction and decreased engagement. Faced with competition from Google, Altman ultimately chose to revert to the more popular GPT-4o model and instructed to reinforce personalization features and the use of user signals again.

Future Outlook: Walking a Tightrope Between Ideal and Reality

OpenAI plans to release another model with improved image quality, speed, and "personality" in January next year, which is expected to end the "Red Code" status. A company spokesperson stated that there is no fundamental conflict between the two philosophies, and the widespread adoption of AI tools is a way to distribute the benefits of AGI to the public This brings to mind the situation faced by social media giant Meta Platforms: wavering between responding to TikTok's competition (launching Reels) and investing in the long-term future (the metaverse). As Jim Steyer, founder of the children's rights advocacy organization Common Sense Media, stated, the social media industry's long-standing prioritization of engagement has led to a mental health crisis, and the current question is whether AI companies will repeat the same mistakes.

For OpenAI, the current challenge lies not only in technological breakthroughs but also in finding a fragile balance between fierce commercial competition, high operating costs, and increasingly sensitive safety ethics