For nearly three years, Marc Benioff, CEO of Salesforce, was a fan of ChatGPT. Then, late last month, he abruptly converted to Google’s chatbot, Gemini. “Holy shit,” he wrote on
When Gemini 3 was released in mid-November, it appeared to crush OpenAI’s flagship model according to a series of reviews shared by Google. The robot has since received widespread praise from the tech industry. One analyst said that Gemini 3 is “the best model ever.» Another crowned Google as the “AI Winners.” Sam Altman seems alarmed: In a company-wide memo last week, OpenAI CEO would have declared a “code red” effort to improve ChatGPT’s capabilities.
OpenAI once had a clear technological advantage. When the company kicked off the AI race in 2022 with the launch of ChatGPT, Google was caught off guard and declared its own “code red”. Google’s first chatbot offerings were indeed a mess: the very first demo of Bard, the precursor to Gemini, included a factual error. A year later, “AI Previews” in Google Search told users that it was healthy to eat a stone a day. Meanwhile, OpenAI has become the most valuable private company in the world, under the assumption that it will always set the tone. But his rise no longer seems inevitable.
OpenAI’s warning lights were flashing even before Google released Gemini 3. OpenAI hasn’t had a stable or even convincing lead on major AI benchmarks for many months. An image generation model released by Google this year, called “Nano Banana”, is significantly faster than ChatGPT and has expanded the user base of Gemini, which in several measurementsis growing several times faster than ChatGPT. Google isn’t the only rival pulling ahead, either: Anthropic’s Claude is widely considered the best model in coding, despite OpenAI’s efforts to catch up. Even Elon Musk’s Grok is roughly on par with the latest version of ChatGPT. (OpenAI, which has a corporate partnership with The Atlanticdid not respond to a request for comment.)
To be fair, this isn’t the first time OpenAI has seemed to lose its edge, only to quickly regain its leadership position in AI. Last year, as Google and Anthropic’s bots seemed to be catching up to ChatGPT, OpenAI released its “reasoning” models and launched an entirely new “reasoning” model. new paradigm of the development of AI. Now, virtually all top AI labs have these “reasoning” models (Gemini 3 is one). In January, when the Chinese AI start-up Deep search After developing a bot equal to and cheaper than those of many major US companies, OpenAI responded with its own new, extremely profitable AI model. This time too, OpenAI could very well make a comeback: its research director, Mark Chen, recently declared at a press conference. podcast that the company has internal models comparable to Gemini 3 which will be released soon. But the company has never seemed so behind in so many areas. More than ever, OpenAI looks like just another chatbot company.
Either way, OpenAI doesn’t really seem focused on creating the “smartest” bot. Instead, the company moved aggressively to establish a business empire. In recent months, OpenAI has rolled out new shopping features, a web browser, an AI-centric social media app, and to top it all off, group chats. Such tools are not exactly stepping stones on the path to digital superintelligence. Instead, they can be understood as a concerted attempt to build a self-sustaining OpenAI ecosystem. ChatGPT is becoming a one-stop shop for everything you might need to do on the internet: browse, work, email, shop, plan a vacation, share AI-generated content with friends. In his “code red” memo, Altman reportedly stated that some of these commercial projects would no longer be prioritized in favor of ChatGPT.
OpenAI’s commercial plans may have come at a cost. According to a recent investigation by THE New York TimesOpenAI has taken user engagement and retention into account in ChatGPT updates. These changes, in turn, may have made some versions of ChatGPT dangerously obsequious – they appeared to praise and reinforce some users’ darkest and most absurd ideas – and were the subject of several lawsuits against OpenAI, alleging that ChatGPT fueled delusional spirals and in some cases even contributed to suicide. (OpenAI denied allegations in the first lawsuit that ChatGPT plunged a user into a mental health crisis, and is currently reviewing a series of more recent cases.)
OpenAI’s efforts to create a family of services are already the go-to playbook for tech giants like Apple and Google to lock users into their products. In that sense, the company was already catching up. What should worry OpenAI the most when Gemini 3 launches is not the technical prowess of the model but the fact that Google immediately began integrating the bot into its existing ecosystem. Google offers at least seven products, each with 2 billion users; OpenAI has not yet reached 1 billion. Altman’s “code red” statement is a reminder that, despite OpenAI’s unprecedented rise, it remains first and foremost a startup.