Technologies

OpenAI Yanked a ChatGPT Update. Here’s What It Said and Why It Matters

The company says it plans to be more careful when releasing updates in the future.

Recent updates to ChatGPT made the chatbot far too agreeable and OpenAI said Friday it’s taking steps to prevent the issue from happening again.

In a blog post, the company detailed its testing and evaluation process for new models and outlined how the problem with the April 25 update to its GPT-4o model came to be. Essentially, a bunch of changes that individually seemed helpful combined to create a tool that was far too sycophantic and potentially harmful.

How much of a suck-up was it? In some testing earlier this week, we asked about a tendency to be overly sentimental, and ChatGPT laid on the flattery: «Hey, listen up — being sentimental isn’t a weakness; it’s one of your superpowers.» And it was just getting started being fulsome.

«This launch taught us a number of lessons. Even with what we thought were all the right ingredients in place (A/B tests, offline evals, expert reviews), we still missed this important issue,» the company said.

OpenAI rolled back the update this week. To avoid causing new issues, it took about 24 hours to revert the model for everybody.

The concern around sycophancy isn’t just about the enjoyment level of the user experience. It posed a health and safety threat to users that OpenAI’s existing safety checks missed. Any AI model can give questionable advice about topics like mental health but one that is overly flattering can be dangerously deferential or convincing — like whether that investment is a sure thing or how thin you should seek to be.

«One of the biggest lessons is fully recognizing how people have started to use ChatGPT for deeply personal advice — something we didn’t see as much even a year ago,» OpenAI said. «At the time, this wasn’t a primary focus but as AI and society have co-evolved, it’s become clear that we need to treat this use case with great care.»

Sycophantic large language models can reinforce biases and harden beliefs, whether they’re about yourself or others, said Maarten Sap, assistant professor of computer science at Carnegie Mellon University. «[The LLM] can end up emboldening their opinions if these opinions are harmful or if they want to take actions that are harmful to themselves or others.»

(Disclosure: Ziff Davis, CNET’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed on Ziff Davis copyrights in training and operating its AI systems.)

How OpenAI tests models and what’s changing

The company offered some insight into how it tests its models and updates. This was the fifth major update to GPT-4o focused on personality and helpfulness. The changes involved new post-training work or fine-tuning on the existing models, including the rating and evaluation of various responses to prompts to make it more likely to produce those responses that rated more highly.

Prospective model updates are evaluated on their usefulness across a variety of situations, like coding and math, along with specific tests by experts to experience how it behaves in practice. The company also runs safety evaluations to see how it responds to safety, health and other potentially dangerous queries. Finally, OpenAI runs A/B tests with a small number of users to see how it performs in the real world.

The April 25 update performed well in these tests, but some expert testers indicated the personality seemed a bit off. The tests didn’t specifically look at sycophancy, and OpenAI decided to move forward despite the issues raised by testers. Take note, readers: AI companies are in a tail-on-fire hurry, which doesn’t always square well with well thought-out product development.

«Looking back, the qualitative assessments were hinting at something important and we should’ve paid closer attention,» the company said.

Among its takeaways, OpenAI said it needs to treat model behavior issues the same as it would other safety issues — and halt a launch if there are concerns. For some model releases, the company said it would have an opt-in «alpha» phase to get more feedback from users before a broader launch.

Sap said evaluating an LLM based on whether a user likes the response isn’t necessarily going to get you the most honest chatbot. In a recent study, Sap and others found a conflict between the usefulness and truthfulness of a chatbot. He compared it to situations where the truth is not necessarily what people want — think about a car salesperson trying to sell a vehicle.

«The issue here is that they were trusting the users’ thumbs-up/thumbs-down response to the model’s outputs and that has some limitations because people are likely to upvote something that is more sycophantic than others,» he said.

Sap said OpenAI is right to be more critical of quantitative feedback, such as user up/down responses, as they can reinforce biases.

The issue also highlighted the speed at which companies push updates and changes out to existing users, Sap said — an issue that’s not limited to one tech company. «The tech industry has really taken a ‘release it and every user is a beta tester’ approach to things,» he said. Having a process with more testing before updates are pushed to every user can bring these issues to light before they become widespread.

Technologies

Forza Horizon 6 Focuses on Japanese Car Culture: Here’s What to Know

The racing series’ latest edition, out in May, includes stunning views of Mount Fuji.

Microsoft has given racing-game fans a deeper dive into the upcoming Forza Horizon 6, and on Friday announced a May 19 release date for Xbox and PC. Premium Edition buyers will get early access on May 15, but PlayStation 5 owners will have to wait until sometime later in 2026 to play.

The publisher provided a first-look video that shows stunning views of Japan, including cherry blossom-strewn rural roads, ice-enclosed courses, neon-lit Tokyo streets and plenty of views of Mount Fuji.

The game will launch with 550 collectible cars, and Microsoft will display two on the cover: the 2025 GR GT Prototype and the 2025 Toyota Land Cruiser.

Don’t miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.

Microsoft says the open-world game will focus on Japanese car culture, with the player in the role of a new arrival to the country guided by two close friends, motorsports fan Jordy and car builder Mei.

The locations will be divided into districts, ranging from bustling urban areas such as Tokyo’s Shibuya Crossing to suburbs, industrial docklands, and more rural areas.

What to know about Forza Horizon 6

Progress in the game will include unlocking new cars and more advanced races, plus a journal of digital mementos inspired by stamp collecting.

The game features eight player houses across Japan that serve as fast-travel points. And new to the game series is The Estate, a piece of land players can acquire and build on.

Inspired by Japanese car culture, Forza Horizon 6 will add Car Meets that’ll serve as social spaces for players.

The Standard Edition costs $70, the Deluxe Edition costs $100, and the Premium Edition, which unlocks early access and includes several Car Packs, costs $120. The game will also be a Day 1 release on Xbox Game Pass Ultimate, and will be available on Steam and Xbox Cloud as well.

Technologies

TikTok Deal Will Keep It Online in the US, but Your Experience of the App Might Change

TikTok has secured its future by agreeing to split the US app from the global business. But the deal will spark changes to the app’s algorithm.

TikTok has dodged a ban and secured its long-term future in the US by announcing a deal on Friday that will see a joint venture take over US operations of the popular social video app. The deal marks the conclusion of a protracted battle over the app’s continued presence in the US, which dates back to President Donald Trump’s first term in office.

TikTok in the US will now be run by TikTok USDS Joint Venture LLC, which was established by a White House executive order issued in September 2025. At its helm will be CEO Adam Presser, previously the head of operations, who led TikTok’s efforts to ensure that the data of the app’s US users was kept secure. Shou Chew, the CEO of TikTok’s international operations, will serve on the joint venture’s board of directors.

«TikTok USDS Joint Venture’s mandate is to secure US user data, apps and the algorithm through comprehensive data privacy and cybersecurity measures,» the company said in a statement. «It will safeguard the US content ecosystem through robust trust and safety policies and content moderation while ensuring continuous accountability through transparency reporting and third-party certifications.»

The venture has three managing investors — Silver Lake, Oracle and MGX — which each hold a 15% stake. Oracle also will be responsible for protection of US user data and of the freshly retrained algorithm, which will be specific to the US version of the app.

Presidents Trump and Joe Biden raised concerns over a potential national security threat posed by TikTok, because of its Chinese-owned parent company ByteDance, which will retain a 19.9% stake in the new joint venture. During both of his presidential terms, Trump has attempted to ban TikTok, but also delayed the ban’s implementation.

The deal announced on Friday arrived moments before the deadline set by the White House for TikTok to comply with its September executive order. In a post on his social site Truth Social, Trump said he was «so happy to have helped in saving TikTok.»

«I only hope that long into the future I will be remembered by those who use and love TikTok,» said Trump. He also thanked China’s President Xi Jinping for working with the US and approving the deal. «He could have gone the other way, but didn’t, and is appreciated for his decision,» he said.

How TikTok might change for you

TikTok has more than 200 million users in the US, and if you’re one of them, the deal announced on Friday will allow you to continue using the app without the ongoing fear of it being banned.

It also won’t see you cut off from creators in China, or the rest of the world. People in the US will still be able to watch videos from Europe, such as last year’s viral «nothing beats a Jet2 holiday» trend. TikTok users outside of the US will still be able to follow their favorite American creators.

In the TikTok newsroom post, the company addressed interoperability, saying that the deal would «provide US users with a global TikTok experience, ensuring US creators can be discovered and businesses can operate on a global scale.»

Where the experience might change is in the content that is recommended to you. Under the terms of the deal, TikTok’s algorithm will be retrained, tested and updated based on US user data. This will have a knock-on impact on what you see on the platform, according to Kelsey Chickering, principal analyst at Forrester.

«TikTok’s power lies in its content graph — an algorithm that learns from thousands of user signals to deliver hyper‑relevant, highly addictive videos,» said Chickering. «With a US joint venture retraining that algorithm on domestic data, the experience will change — maybe for the better, maybe not. One thing’s certain: TikTok in America won’t be the same.»

In spite of the interoperability that will see US TikTok users connected to those across the globe, it does seem likely that the focus on US data will lead to a shift away from the global nature of the content that the algorithm currently serves up to you.

«TikTok’s US algorithm will now be trained on US data, which means what trends — and what dominates feeds — will feel distinctly American,» said Chickering. «Global content will still appear, but its ranking will change.»

Exactly how this will look may differ from person to person, and will likely take some time to come into effect as the joint venture begins the retraining process. TikTok didn’t immediately respond to questions regarding how long it expects retraining the algorithm to take, when US TikTok users should expect to be impacted by changes and whether it will issue public updates about this process.

One potential pitfall the company might want to avoid, Chickering said, is moderating the US version of TikTok in a way that tilts too far toward any one particular political viewpoint, or fails to curb misinformation. Elon Musk’s takeover of Twitter (now X) — and his subsequent algorithmic changes that alienated users and advertisers — is a cautionary tale in this regard. With Instagram Reels already vying to replace TikTok, the company will likely want to avoid making changes that could spark a mass exodus of people.

«For now, it’s speculation,» said Chickering. «It remains to be seen how new leadership will wield this power and whether moderation policies will evolve.»

Technologies

Today’s NYT Mini Crossword Answers for Friday, Jan. 23

Here are the answers for The New York Times Mini Crossword for Jan. 23.

Looking for the most recent Mini Crossword answer? Click here for today’s Mini Crossword hints, as well as our daily answers and hints for The New York Times Wordle, Strands, Connections and Connections: Sports Edition puzzles.

Need some help with today’s Mini Crossword? Hope you’re familiar with a certain blond actor (8-Across)! Read on for all the answers. And if you could use some hints and guidance for daily solving, check out our Mini Crossword tips.

If you’re looking for today’s Wordle, Connections, Connections: Sports Edition and Strands answers, you can visit CNET’s NYT puzzle hints page.

Read more: Tips and Tricks for Solving The New York Times Mini Crossword

Let’s get to those Mini Crossword clues and answers.