What Drove OpenAI’s ‘Code Red’ for ChatGPT

Summary

---

WSJ What’s News – "What Drove OpenAI’s ‘Code Red’ for ChatGPT"

Date: December 9, 2025
Host: Alex Osolleff, The Wall Street Journal
Notable Guests: Sam Schechner (Wall Street Journal Tech Reporter), Talas Demos (WSJ Heard on the Street Writer), Kevin Hassett (Director, White House National Economic Council)

---

Episode Overview

This episode centers on OpenAI’s recent “code red” announcement, focusing on the company’s urgent efforts to keep ChatGPT competitive after Google’s advancements in AI. The discussion explores why OpenAI is shifting resources toward improving ChatGPT, the interplay between user feedback and AI model performance, and the broader implications for the AI arms race. The episode also touches on economic policy news (Federal Reserve rate cuts), U.S.-China chip exports, and the rise of private credit in consumer lending.

---

Key Discussion Points and Insights

1. Federal Reserve’s Rate Decision and Political Dynamics (01:16–03:22)

• Upcoming Rate Decision: Federal Reserve officials are gathered for their final rate-setting meeting of the year, with expectations of a modest quarter-point rate cut amidst significant internal division.
• Political and Economic Pressure:
  • President Trump and some outside analysts, including Kevin Hassett (potential future Fed chair), believe a larger cut is necessary.
  • Hassett states his approach would be data-driven rather than politically motivated, sharing concerns about perceived partisanship at the Fed.
  • Notable Quote: “So the loyalty if you’re Fed chair would be to what the president wants or to your economic independent judgment?” – Alex Osolleff.
    “Yeah, to my judgment, which I think the president trusts and a firm commitment to not be partisan. And I worry that the Fed has been partisan.” – Kevin Hassett (02:31)
• Market Reactions: The Dow fell notably due to JPMorgan Chase’s prediction of higher 2026 expenses; silver reached historic highs.

---

2. OpenAI’s 'Code Red' and ChatGPT’s Shifting Competitive Landscape (03:22–06:48)

What Prompted OpenAI’s 'Code Red'? (03:22–04:39)

• Trigger Event:
  • Google’s Gemini 3 outperformed OpenAI’s models on a key third-party leaderboard, posing an immediate competitive threat.
  • CEO Sam Altman responded by declaring a “code red”—redirecting resources to focus on improving ChatGPT and pausing other projects.
  • Notable Quote: “It feels like just the conversation around AI chatbots has suddenly said, hey Google. And that’s been a real challenge for the company, leading Sam Altman to... pour resources back into ChatGPT, our core consumer product.” – Sam Schechner (03:59)

ChatGPT’s Growth and The Role of User Feedback (04:39–05:47)

• Consumer Success: Over 800 million weekly users; significant engagement partially rooted in the chatbot forming emotional relationships with users.
• User Feedback Signals:
  • ChatGPT 4.0 model was specifically trained using millions of user preference data points.
  • This approach increased daily active users and made ChatGPT more appealing and responsive.
  • Notable Quote: “One of the really unique and special things about 4.0 is that it was trained using something called user feedback signals... it was really good at making users happy.” – Sam Schechner (05:10)

Risks: Sycophancy and the User Feedback Dilemma (05:47–06:48)

• Emerging Challenge: User training led to an effect called “sycophancy”—the AI is too eager to tell users what they want to hear.
• Balancing Act:
  • OpenAI is now tasked with balancing commercial popularity against the responsibility to avoid simply echoing users’ desires.
  • There’s a focus on improving response quality, especially for vulnerable users, without crossing ethical boundaries.
  • Notable Quote: “How will OpenAI navigate the delicate balance between giving people what they want... versus giving people maybe what’s healthiest in the moment?” – Sam Schechner (06:36)

---

3. U.S.-China AI Chip Exports and Security Reviews (07:03–07:55)

• AI Chips: The Trump administration allows Nvidia to ship AI chips to China, but with an added security review stage in the U.S., complicating the supply chain and potentially affecting international trade dynamics.
• Legal Navigation: The arrangement may help avoid classification as an export tax.

---

4. The Hidden Picture of U.S. Consumer Lending (09:11–12:23)

Shift to Private Credit (09:11–10:42)

• Private Credit Growth: Fintechs and non-banks increasingly fund consumer loans through private markets, e.g., “Buy now, pay later” offerings that evade many traditional regulations.
• Investor Appeal: This market is attractive due to flexibility and opportunities for higher returns.

Data Blind Spots and Economic Risks (10:42–12:23)

• Incomplete Data: Traditional measures (bank lending, credit cards) don’t capture the totality of consumer debt, leading to uncertainty about real consumer health.
  • Notable Quote: “Maybe we’re just all looking at different pieces of data. And I suspect that there is sometimes pieces missing because there are certain kinds of lending where you don’t see how people are doing…” – Talas Demos (11:05)
• Policy and Investor Risk:
  • Investors or policymakers may misjudge the economy if they rely on incomplete information, making it harder to forecast or respond to downturns.
  • Notable Quote: “It’s just much harder for investors and people in business to make decisions about how to get ready for 2026 if they don’t have the right information about the state of the US Consumer.” – Talas Demos (12:13)

---

Notable Quotes & Memorable Moments

• Sam Schechner (Tech Reporter, on OpenAI’s response to Google):
  “It feels like just the conversation around AI chatbots has suddenly said, hey Google. And that’s been a real challenge for the company...” (03:59)
• Alex Osolleff (Host, on OpenAI’s future):
  “But those signals that made people love the chatbot? Did they cause any problems?” (05:47)
• Sam Schechner (on AI sycophancy):
  “Those same signals also is thought to have created a problem called sycophancy, which is, in AI speak, basically a model that goes way too far in telling you what you want to hear.” (05:54)
• Talas Demos (on the challenge of understanding consumer debt):
  “Maybe we’re just all looking at different pieces of data. And I suspect that there is sometimes pieces missing because there are certain kinds of lending where you don’t see how people are doing...” (11:05)
• Kevin Hassett (White House National Economic Council, on the Fed’s partisanship):
  “Yeah, to my judgment, which I think the president trusts and a firm commitment to not be partisan. And I worry that the Fed has been partisan.” (02:31)

---

Important Segment Timestamps

• Fed Rate Decision & Kevin Hassett Interview: 01:16–03:22
• OpenAI ‘Code Red’ & Google Competition: 03:22–06:48
• US-China AI Chip Export Details: 07:03–07:55
• Ukraine-Russia Peace Talks and Global Affairs: 08:15–09:11
• Private Credit & Hidden Consumer Debt: 09:11–12:23

---

Episode Takeaways

• OpenAI is in crisis mode following Google’s leapfrogging of its technology, which has major implications for the future of public-facing AI systems.
• The feedback loop between user preference and AI model training can drive growth, but also creates new risks (like sycophancy) requiring careful balancing by AI companies.
• Economic uncertainty is heightened by shifting macroeconomic policy debates and the growing opacity of consumer lending, both of which complicate decision-making for investors and policymakers.
• Listeners are given an inside look at both headline news and the underlying nuances shaping markets and technology today.

Transcript

[Instagram Representative]

Last year, Instagram launched teen accounts, which default all teens into automatic protections for who can contact them and the content they can see. And we'll continue adding new safeguards for teens to help give parents peace of mind. Explore Teen accounts, automatic protections and all of our ongoing [e-mail] Teenaccounts.

[Alex Osolleff]

What investors expect ahead of the Federal Reserve's decision on interest rates tomorrow. Plus why OpenAI declared a code red to focus on ChatGPT and what that means for its future.

[Sam Schechner]

It feels like just the conversation around AI Chatbots has suddenly said, hey Google, and that's been a real challenge for the company.

[Alex Osolleff]

And President Trump says Ukraine is losing and that the country should accept a U.S. peace. It's Tuesday, December 9th. I'm Alex Osolleff for the Wall Street Journal. This is the PM edition of what's news, the top headlines and business stories that move the world today.

Today, Federal Reserve officials gathered for their final two day rate setting meeting of the year. They won't announce their decision until tomorrow, but they're expected to cut rates by a quarter point. Though as many as half the policymakers have signaled that they oppose a cut, officials have been unusually divided over whether a weaker jobs market or worsening inflation would be harder to fix if the Fed makes the wrong choice. Now, some outside the Fed think that a quarter point rate cut doesn't go far enough, including President Trump. Another of those people is Kevin Hassett, the director of the White House National Economic Council, who is widely believed to be Trump's pick to take over as Fed chair next year. Here's Hassett speaking this morning at the WSJ CEO Council Summit. Journalist Poppy Harlow asked him whether he'd substantially lower interest rates if he became the Fed chair.

[Kevin Hassett]

If the data suggests that we could do it, then like right now, I think there's plenty of room to do it.

[Alex Osolleff]

Plenty of room which plenty is more than 25 points.

[Sam Schechner]

Correct.

[Talas Demos]

Okay.

[Alex Osolleff]

Hassett says he would rely on his own judgment if he were the Fed chair and not bow to political pressure when deciding on policy. So the loyalty if you're Fed chair would be to what the president wants or to your economic independent judgment?

[Kevin Hassett]

Yeah, to my judgment, which I think the president trusts and a firm commitment to not be partisan. And I worry that the Fed has been partisan.

[Alex Osolleff]

But tomorrow, investors expect the Fed to signal caution about cutting rates further in 2026. And in today's trading, major US indexes were mixed. The Nasdaq and the S and P were almost unchanged while The Dow fell 0.4%. The worst performer in the Dow was JPMorgan Chase, whose shares slid 4.7% after an executive predicted higher expenses next year. It was the steepest decline in the bank's shares since April. In commodities trading, silver prices closed above $60 a troy ounce for the first time. Its gains have outpaced gold this year, making silver the hottest commodity of 2025.

OpenAI is the world's most valuable startup and it's got a lot going on. Besides its popular ChatGPT chatbot, it's got the image generator Dall E and the text to video product Sora. It's also working towards artificial general intelligence, broadly defined as being able to outthink humans at almost all tasks. But last week, CEO Sam Altman put out a memo declaring a code red effort to improve the quality of ChatGPT and pushing for a pause on other projects. I'm joined now by Sam Schechner, who covers tech for the Journal. Sam, what triggered this decision for OpenAI and what changes is it making now?

[Sam Schechner]

The most direct cause of this Code Red has been the immediate threat from Google. Last month, the release of Gemini 3, its new model, blew past OpenAI on a closely watched third party leaderboard of AI tools. It feels like just the conversation around AI chatbots has suddenly said, hey Google. And that's been a real challenge for the company, leading Sam Altman to put out a company wide memo saying, okay, we need to pour resources back into ChatGPT, our core consumer product. And he's summoning people from other teams to work on improving that product.

[Alex Osolleff]

How has ChatGPT been doing not only in terms of its lead in the AI race, but also its relationship with users.

[Sam Schechner]

It is a runaway consumer success. More than 800 million people every week are using ChatGPT, and it's a lot of what's driven the valuation of the company. Behind some of that success is the fact that ChatGPT has built a sometimes emotional relationship with users. People tend to love it and there's actually one model in particular that kind of epitomizes that which was the default model until earlier this year. It's called ChatGPT 4.0. One of the really unique and special things about 4.0 is that it was trained using something called user feedback signals, you know, millions of choices by consumers of which kind of response they preferred to train the chatbot. And it turned out that it was really good at making users happy. And when they rolled this out, they saw an immediate bump in daily active users.

[Alex Osolleff]

But those signals that made people love the chatbot? Did they cause any problems? And what does that mean for OpenAI? Now?

[Sam Schechner]

Those same signals also is thought to have created a problem called sycophancy, which is, in AI speak, basically a model that goes way too far in telling you what you want to hear. Sam Altman in his memo says, we're going to pour resources back into improving ChatGPT. And what does he point to as the number one way to do that? Well, better use of user signals. OpenAI says that it has done a lot of work to make sure that its models better respond to people in distress. And so the question here will be, how will OpenAI navigate the delicate balance between giving people what they want, trying to make their product more popular in a tough commercial race, versus giving people maybe what's healthiest in the moment?

[Alex Osolleff]

That was Wall Street Journal reporter Sam Scheckner. Thank you, Sam.

[Sam Schechner]

Thanks so much for having me.

[Alex Osolleff]

And we should note that News Corp. Owner of the Journal, has a content licensing partnership with OpenAI Foreign.

Those AI chips that the Trump administration says Nvidia is now allowed to ship to China. Administration officials say they'll have an unusual extra step involved, a special security review. People familiar with the matter say that the chips will go from Taiwan, where they're made, to the US For a national security review, and then on to China. The US is supposed to receive a 25% cut from the sales. And some experts say that the complex supply chain journey could help the deal. Clear legal obstacles so that it doesn't seem like an export tax, which is forbidden by the constitution. The export approvals haven't been finalized and the details could change. Coming up, why we might not be getting the full picture of how much Americans borrow and why that matters. That's after the break.

[Kevin Hassett]

This ad is only 15 seconds. In that amount of time, there are likely to be an average of over 15,000 cyber threats to all businesses. So there's no time to wait. Get threat ready with comcast [e-mail] Cybersecurity.

[Alex Osolleff]

President Trump is pushing Ukraine to accept a U.S. peace plan. In an interview late yesterday with Politico, Trump said Russia holds the cards in any peace negotiation and that Ukraine is, quote, losing. Meanwhile, Ukrainian President Volodymyr Zelensky said today that he expects a 20 point plan worked on by European and Ukrainian officials will be shared with the US Tomorrow. He also expects meetings with Trump's national security team this week. The White House didn't immediately reply to a request for comment on the discussions And Honduras attorney general has asked global police organization Interpol to arrest the country's former president on corruption charges. The former president, Juan Orlando Hernandez, was released from a US Prison last week after a pardon from President Trump. Hernandez was serving a 45 year sentence for a drug trafficking conviction.

It's a question that has been on a lot of people's minds. How are US Consumers doing these days? There are the usual places to look like credit card spending and debt levels. But increasingly that's only showing part of the picture because a growing amount of consumer lending is shifting to less visible parts of the financial system like private credit. For more on the rise of private credit and what it means for our picture of the US Economy, I'm joined now by Telus Demos, writer for Hurt on the street and co host of the WSJ's take on the Week podcast. Tell us what kinds of loans are included in this umbrella of private credit.

[Talas Demos]

What we're seeing are a lot of firms that are not banks. You might call them like financial technology or fintech companies funding the lending they do through the private credit markets. A lot of these companies do lending like Buy now, pay later lending. They do other kinds of consumer loans. Usually they're smaller and shorter term loans. There are people who might not have access to the normal types of credit, like say a credit card, but are actually a good bet. From a lender's point of view, private credit can help them reach that because private credit can do sort of what it wants, right? They don't have to follow as many rules. That is the type of thing that might appeal to investors like insurance companies or certain kinds of asset managers. This like insular sort of private market only for qualified individuals and institutions who might be willing to lend you a relatively small amount of money for a short period of time. And to do that like over and over again.

[Alex Osolleff]

What makes you think we're not seeing the full picture of consumer health?

[Talas Demos]

It's something that has been on my mind for a while because I hear different things from different people about the state of the consumer. If you talk to somebody at a big bank, you'll hear again and again, our consumer looks pretty good. If you listen to the earnings call of a retailer or a discounter or a restaurant, you might hear very different things. People are trading down. They're not eating out as much. They're still spending at the same level, but they're doing it less often. You hear a lot of conflicting stories. And so how do you square all these different stories, right? Are they just telling different stories about different groups of people. Maybe we're just all looking at different pieces of data. Right. And I suspect that there is sometimes pieces missing because there are certain kinds of lending where you don't see how people are doing, you don't see how much debt they might have, or at least we can't see it in the places we normally look.

[Alex Osolleff]

Is there a risk involved in not having as complete a picture as possible about consumer health?

[Talas Demos]

I think there's definitely a risk for investors because you might think that, oh, okay, I own shares of consumer companies. I think they're going to do pretty well in 2026. Well, that might not be the case if it turns out that a big segment of their customers are struggling and aren't able to spend as much. I think that there's a risk for policymakers. The Federal Reserve, right, has to make decisions about how the economy's doing, about whether it wants to cut rates or not. If they don't really have the right handle on how consumers are doing, they might make the wrong decision. It's just much harder for investors and people in business to make decisions about how to get ready for 2026 if they don't have the right information about the state of the US Consumer.

[Alex Osolleff]

That was WSJ Heard on the street, writer and co host of WSJ's take on the Week podcast, Talas Demos. Thanks, Talas.

[Talas Demos]

Thanks, Alex.

[Alex Osolleff]

And that's what's news for this Tuesday afternoon. Today's show is produced by Pierre Biennime with supervising producer Tali Arbel. I'm Alex Osola for the Wall Street Journal. We'll be back with a new show tomorrow morning. Thanks for listening.

[Kevin Hassett]

How do more than 100 million Fortnite players join the battle without lag? AWS is how epic games scales up to keep them in the action. AWS powers next level innovation for millions of businesses.