Highlights
- AI is confusing — here’s your cheat sheet
- Figma explains how its AI tool ripped off Apple’s design
- Early Apple tech bloggers are shocked to find their name and work have been AI-zombified
- Google and OpenAI are racing to rewire the internet
- We have to stop ignoring AI’s hallucination problem
- OpenAI releases GPT-4o, a faster model that’s free for all ChatGPT users
- The true promise of AI: Siri that doesn’t suck.
- Zuckerberg says it will take Meta years to make money from generative AI
- Meta’s battle with ChatGPT begins now
- Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis
- One month with Microsoft’s AI vision of the future: Copilot Pro
- How much electricity does AI consume?
- How AI copyright lawsuits could make the whole industry go extinct
- Gemini Advanced is most impressive when it’s working with Google
- Google’s AI now goes by a new name: Gemini
- ChatGPT is winning the future — but what future is that?
- OpenAI is letting anyone create their own version of ChatGPT
- Microsoft CTO Kevin Scott thinks Sydney might make a comeback
- AI chatbots compared: Bard vs. Bing vs. ChatGPT
- OpenAI announces GPT-4 — the next generation of its AI language model

You can’t trust your eyes to tell you what’s real anymore, says the head of Instagram


Cath Virginia / The Verge | Photos from Getty Images
Instagram boss Adam Mosseri is closing out 2025 with a 20-images-deep dive into what a new era of “infinite synthetic content” means as it all becomes harder and harder to distinguish from reality, and the old, more personal Instagram feed that he says has been “dead” for years. Last year, The Verge’s Sarah Jeong wrote that “...the default assumption about a photo is about to become that it’s faked, because creating realistic and believable fake photos is now trivial to do,” and Mosseri eventually concurs:
You can read the full text from his slideshow at the bottom of this post, but according to Mosseri, the evolution required of Instagram and other platforms is that “We need to build the best creative tools. Label AI-generated content and verify authentic content. Surface credibility signals about who’s posting. Continue to improve ranking for originality.”

I’m obsessed with Redfin’s AI search


Look, I’m as fed up as the next guy with AI chatbots stuffed into every app. I don’t want to brainstorm coverage options with an LLM every time I renew my car insurance. I’d much rather message a human than a robot to pester FedEx about my missing package. But I have found one scenario where AI is actually pretty great: real estate.
I need to confess something: I’m a Redfin looky-loo. A Zillow zealot. Not because I am actually shopping for a new home. With these interest rates? God, no. But I am perpetually window-shopping for a new home — partly out of nosiness, and partly because I like imagining what life might look like in a different arrangement of bedrooms and bathrooms.

Spotify says it’s working with labels on ‘responsible’ AI music tools


Cath Virginia / The Verge | Photo from Getty Image
The recently-rumored AI partnership between Spotify and the major record labels is now a reality. The streaming service announced today that it’s entering into an agreement with Sony Music Group, Universal Music Group, Warner Music Group, Merlin, and Believe to develop “responsible AI products.” Unfortunately, your guess is as good as ours as to what exactly that means.
Spotify didn’t detail any specific products in the works but said it was building a “state-of-the-art generative AI research lab and product team focused on developing technologies that reflect our principles and create breakthrough experiences for fans and artists.”

I’ve fallen into Sora’s slippery slop


An anime version of Jesus Christ flipping tables. OpenAI employees performing in Hamilton costumes. News anchors discussing a story on television. A man doing a thirst-trap TikTok dance. Sam Altman — stealing GPUs on CCTV, listening to a business pitch, crying.
Such were the contents of my feed on Sora, OpenAI’s new social media app for AI-generated video. The company released the iOS app on Tuesday with the ability to create 10-second videos of virtually anything you can dream up, including “cameos,” or videos featuring your own AI-generated self and anyone else who approves of you using their likeness. OpenAI employees called Sora a potential “ChatGPT moment for video generation” in a briefing with reporters earlier this week. On Friday, Sora topped the list for top free apps in Apple’s App Store.

What happens when an AI-generated artist gets a record deal? A copyright mess


Image: Cath Virginia / The Verge, Xania Monet
Two weeks ago, record company Hallwood Media signed a deal with Telisha “Nikki” Jones after negotiations that purportedly included an offer of $3 million, Billboard reported. Jones is a Mississippi-based lyricist behind the R&B artist “Xania Monet” whose most popular song on Spotify racked up over 1 million listens, and whose Reels regularly top 100,000 views on Instagram – despite her likeness, vocals, and music being AI-generated.
Multiple copyright experts speaking with The Verge have been quite clear: the law is not at all settled but generally one cannot copyright AI-generated works by themselves without human intervention, but you may be able to secure copyright where there are human-made expressive elements, which in this case are the lyrics. So, what exactly is Hallwood Media buying? What can they license? What does this mean for the future of music as a sellable product? The more questions we asked, the more it became evident that we’re facing a cultural shift in the wake of the flood of AI-generated content. The law is just trying to keep up.

New YouTube AI tools help creators give viewers what they want


Influencers and content creators are many things beyond their public personas. All but the biggest figures likely do some combination of the following jobs themselves: content moderator, video editor, photographer, social media strategist, script writer, and idea generator. What if they could outsource much of that work to AI? And what if it were the social media platforms themselves that provided them with the tools to do so?
At the Made on YouTube event, held Tuesday in New York, the company launched a slew of new AI features aimed at content creators, many of which focus on all the behind-the-scenes work that goes into a video. Unlike previous tools — like an AI background music generator or tools that create AI photos and videos — the new ones are largely content strategy features, marketed as helping creators reach new audiences (and more effectively get in front of their existing base).

Making ChatGPT less annoying.
OpenAI chief Sam Altman said the company will be rolling out updates to the chatbot’s personalization page in the “next couple of days.” The update will bring previously disconnected features like “custom instructions” and communication preferences that can make life easier under one heading.

AI chatbots can help perfect a phishing scam, despite being trained not to.
Six of the major AI chatbots - Grok, ChatGPT, Meta AI, Claude, DeepSeek, and Gemini - effectively guided a team of Reuters reporters through the steps of simulating a phishing scam, down to describing a good time to send a message intended to trick older adults into clicking on a fraudulent link.

Silicon Valley’s most powerful alliance just got stronger


Getty Images / The Verge
Eddy Cue deserves a raise.
As the executive overseeing Apple’s services division, he’s highly incentivized to protect the tens of billions of dollars a year that Google pays to be the default search engine in Safari. “I’ve lost a lot of sleep thinking about it,” he said from the witness stand during Google’s antitrust trial earlier this year.

Is AI the end of software engineering or the next step in its evolution?


Image: Cath Virginia / The Verge
The first time I used ChatGPT to code, back in early 2023, I was reminded of “The Monkey’s Paw,” a classic horror story about an accursed talisman that grants wishes, but always by the most malevolent path — the desired outcome arrives after exacting a brutal cost elsewhere first. With the same humorless literalness, ChatGPT would implement the change I’d asked for, while also scrambling dozens of unrelated lines. The output was typically over-engineered, often barnacled with irrelevant fragments of code. There were some usable lines in the mix, but untangling the mess felt like a detour.
When I started using AI-assisted tools earlier this year, I felt decisively outmatched. The experience was like pair-programming with a savant intern — competent yet oddly deferential, still a tad too eager to please and make sweeping changes at my command. But when tasked with more localized changes, it nailed the job with enviable efficiency.

An art critic vs. the White House’s ‘weird AI paintings.’
In general, when one of the Founding Fathers did something that fits contemporary standards of equality, such as speak out for the rights of women or enslaved Africans, their AI avatar mentions it. Anything that is more controversial about them is downplayed or passed over in silence.

“Delusional risk score: near zero.”
The WSJ found that note in a “Clinical Cognitive Profile” ChatGPT provided to Stein-Erik Soelberg, a “56-year-old tech industry veteran with a history of mental instability,” who killed his mother and took his own life earlier this month.
According to the Journal, “...ChatGPT treated his ideas as genius and built upon his paranoia.”

Google is upgrading the Gemini app’s AI image editor.
Google says the Gemini app is better at performing more advanced image editing requests, such as blending two photos, performing multiple edits on the same image, combining different designs, or giving you an outfit change while preserving your appearance.

The head of ChatGPT was “surprised” by how much people were attached to GPT-4o.

The head of ChatGPT on AI attachment, ads, and what’s next


Image: The Verge / Photo: OpenAI
Welcome to Decoder! I’m Alex Heath, your Thursday episode guest host and deputy editor at The Verge. Today, I’m talking to a very special guest: Nick Turley, the head of ChatGPT at OpenAI.
While Sam Altman is definitely the public face of the company, Nick has been leading ChatGPT’s development since the very beginning. It’s now the fastest-growing software product of all time, reaching 700 million people each week.

ChatGPT won’t remove old models without warning after GPT-5 backlash


Image: The Verge
After the backlash to replacing its 4o model with GPT-5, OpenAI will no longer get rid of old models without a heads up.
“In retrospect, not continuing to offer 4o, at least in the interim, was a miss,” Nick Turley, OpenAI’s head of ChatGPT, said on Tuesday. In an interview with me for an upcoming episode of Decoder, he said it was surprising to see the “level of attachment” people had to 4o. “It’s not just change that is difficult for folks, it’s also the fact that people can have such a strong feeling about the personality of a model.”

Quick fixes: bring back Google Photos classic search instead of ‘Ask Photos’ Gemini AI


Image: Cath Virginia / The Verge
You just want to search your archived photos, instead of interacting with an AI assistant.
Inside the app, tap the Google account button at top-right, then go to Photos settings > Preferences > Gemini features in Photos, and turn off either “Search with Ask Photos” or “Use Gemini in Photos”.

What is it really like to go off the rails with ChatGPT?
This article digs into the ChatGPT conversation history of a 47-year-old Canadian man who was told by the bot that he’d “discovered a novel mathematical formula” that could take down the internet and create a force field vest.
When he realized this was delusional, he checked by asking Google Gemini.

GPT-5 is being released to all ChatGPT users


OpenAI is releasing GPT-5, its new flagship model, to all of its ChatGPT users and developers.
CEO Sam Altman says GPT-5 is a dramatic leap from OpenAI’s previous models. He compares it to “something that I just don’t wanna ever have to go back from,” like the first iPhone with a Retina display.

Google swears it isn’t destroying the web with AI search


Image: Cath Virginia / The Verge
Google says its AI search features aren’t tanking website traffic despite recent reports suggesting otherwise. In a blog post published on Wednesday, Google Search head Liz Reid says click volume from the search engine has remained “relatively stable” when compared to the same time last year — though some types of websites are getting more clicks and others are getting less.
Reid’s argument that AI is actually helping the web comes just weeks after Pew Research published a report that said people are “less likely” to click on links when Google presents them with an AI Overview, or the AI-generated summary that appears at the top of Search. In her post, Reid says Google’s findings contradict third-party reports that are “often based on flawed methodologies.”

Mark Zuckerberg promises you can trust him with superintelligent AI


Image: Cath Virginia / The Verge; Getty Images
Hours before Meta’s earnings call, CEO Mark Zuckerberg shared his vision for the future of AI: personalized super-smart AI for everyone — especially in the form of wearable glasses.
He said his vision is for everyone to have an AI tool that “helps you achieve your goals, create what you want to see in the world, experience any adventure, be a better friend to those you care about, and grow to become the person you aspire to be.”

Google rethinks search results with its new AI-curated ‘Web Guide’


Image: Google
On Thursday, Google launched Web Guide, a new AI-powered search feature that “groups links in helpful ways” for people who opt in to the test on Labs. Web Guide runs on a custom version of Google’s Gemini AI model to process search queries, reorganizing the traditional “10 blue links” by sorting results and finding related questions using generative AI.
For anyone who chooses to enable Web Guide, it is currently available within the same “Web” tab on Google Search that currently shows search results without including AI Overviews. It uses the same “query fanout technique” as AI Mode, allowing Gemini to break down queries and make several searches at once.

A look at the state of deepfakes.
If you’re interested in a high-level overview of the state of deepfakes, I recommend checking out this blog post from Captions, a Capcut competitor and a software platform that enables creators to generate AI-generated videos from scratch.
What struck me the most is the company’s prediction that, “Very soon, most models will allow you to generate a person (real or synthetic) in any situation – without a duration constraint (longer than 12 seconds) and featuring multiple people in one shot.” Chat, are we cooked?
I’ll have Captions CEO Gaurav Misra on Decoder later this week to talk about this.

Why AI is moving from chatbots to the browser


Image: The Verge
Happy Friday. I’m back from vacation and still getting caught up on everything I missed. AI researchers moving jobs is getting covered like NBA trades now, apparently.
Before I get into this week’s issue, I want to make sure you check out my interview with Perplexity CEO Aravind Srinivas on Decoder this week. It’s a good deep dive on the main topic of today’s newsletter. Keep reading for a scoop on Substack and more from this week in AI news.

If you can’t bring the AI chips to you...
Then bring suitcases full of hard drives to the AI chips. That’s what some Chinese engineers have reportedly taken to amid efforts to skirt the US ban on selling training chips to China, manually moving terabytes of data to Malaysia to build an AI model there.

