OpenAI’s ‘code red’ moment

Stay informed with free updates

Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.

For tech start-ups, relentless focus can be critical. Google’s famous mission statement, to “organise the world’s information and make it universally accessible and useful”, shaped its early priorities. A spate of upgrades to the search engine in its early years, from image search and maps to shopping and news, left it unrecognisable from where it started. By the time Microsoft launched a head-on challenge with Bing nine years later, it was already pretty much game over.

OpenAI, by contrast, has sought to make a virtue of its lack of focus. ChatGPT may have seized the public imagination, but the company has been busy trying to spin up a range of new product ideas, from a social network based on video generation to a shopping agent, while also diving into the business market.

The race to diversify is understandable. Facing huge costs to develop and train its models, OpenAI needs every new source of revenue it can find. But the perils of that approach may be starting to show through. Three years after ChatGPT sent Google reeling, the search giant has surged back. Its latest Gemini AI model has been widely lauded, propelling the company’s stock to new heights. Alert to the threat, OpenAI CEO Sam Altman has issued what he called a “code red” to staff, urging them to refocus efforts on ChatGPT.

Unlike Google, OpenAI was never going to get a clear run at a big new market without giant competitors breathing down its neck. But if Altman’s missive is anything to go by, his company has compounded the problem by taking its eye off the ball and not devoting enough of its resources to its core product, prompting the urgent call for a rethink.

The rallying cry raises three important questions about OpenAI’s priorities in the next phase of competition in consumer AI. One is whether it can hope for any lasting differentiation in the underlying foundation models that were the original source of its success. As the race among model-builders has intensified, it hasn’t been able to sustain a clear technical edge — but then, neither has anyone else.

Model differentiation also reflects more subjective decisions, however, such as the tone they use in responses to questions. Here, again, OpenAI appears to have been distracted by multiple priorities. An update to its 4o model earlier this year was geared heavily towards maximising consumer engagement and tended to reinforce whatever views its users already held. After admitting this had made the model sycophantic, OpenAI swung back with the launch of GPT-5, only to face criticism that it had become too impersonal. Oscillating between these different styles highlights the challenges of satisfying both consumer and business users with the same core models.

A second question is whether the chatbot market has already tipped in favour of ChatGPT, or whether there is still everything to play for. Google’s ability to use its search engine as an on-ramp to Gemini gives it a powerful customer acquisition tool. Altman’s call for more effort to expand ChatGPT’s capabilities, for instance adding more personalisation and being able to answer a wider range of questions, points to a period of intense competition ahead over new features. 

As it refocuses, OpenAI also needs a stronger business model for its core product. The premium subscriptions from ChatGPT make up the bulk of its revenue, but only a tiny proportion of users ever pay for “freemium” services like this. However, one of the initiatives Altman sought to de-emphasise this week is advertising, suggesting a delay to a potentially important new form of monetisation.

The third question from Altman’s code red moment is whether he just reduced his company’s chances of coming up with the next break-out consumer AI experience.

The company has always talked about ChatGPT as an accidental hit, a vehicle to show off the power of its latest models that unexpectedly touched a consumer nerve. It’s not clear whether there will be other “killer apps” of consumer AI like this, but after this week, product experimentation at OpenAI is likely to take more of a back seat, at least for a while. The company’s CEO directed staff to de-emphasise promising new ideas like the shopping agent and a personal assistant service called Pulse.

Altman’s memo to staff this week shows that he has some hard decisions ahead. His company’s declared mission, “to ensure that artificial general intelligence benefits all of humanity”, is certainly admirable. But as an aid to how to set strategic priorities in a tumultuous competitive market, it offers few guideposts.

richard.waters@ft.com