Last year, Google DeepMind sent gaming stocks tumbling with the reveal of its Genie 3 world model. Since then, world models have become a major focus of AI labs around the world, but Google isn’t resting on its laurels. Researchers have been hard at work on an updated model that improves on Genie 3 in almost every way. And if you pay Google enough money, you might actually get to use Genie 4 before too long.
It’s easy to look at the output of Genie 4 and see a video game, which is why Genie 3 caused so much consternation in the gaming industry. DeepMind doesn’t think about it that way, though. According to Google, world models are a key element in the march toward artificial general intelligence (AGI). A world model creates a virtual world, but it doesn’t require pre-rendering environments or programming interactive elements. You simply tell the model what you want, and it generates the experience frame by frame in real time as you explore it.
DeepMind says this technology could create infinite training environments for physical robots or unlock new avenues of rapid prototyping for creative fields. Of course, that requires the model to accurately simulate what you want. Genie 4 has several improvements that make its simulations more useful, and that’s thanks in part to Google’s updated Gemini models.
When you prompt a world into existence in Genie 4, it leverages Gemini 3.8 Flash as a guiding agent to break down goals and evaluate each output frame. This helps keep the simulation on track, letting you move through varying environments without losing the original context. However, DeepMind notes that Genie 4 worlds still only remain consistent for a couple of minutes before they forget how things started.
Like Genie 3, the new version supports changing the parameters midstream. You can change the weather, add elements, or move the camera, and Gemini will help steer the simulation with each update. This is similar to the way reasoning models generate additional internal prompts to zero in on the right output tokens.
The physics in AI-generated video can sometimes look a bit unnatural, but DeepMind says Genie 4 has reached a new level of realism. Genie 4 can apparently do a better job with elements like aerodynamics and collisions, which makes the simulations more useful and visually appealing. People appearing in these simulations should also look less uncanny, with more natural movement and responses to the changing environment.
Genie 4 is also the first DeepMind world model with native audio. Videos will have ambient background sounds, and you can specify audio elements in your prompt. The biggest leap, if it works, is the addition of human speech. DeepMind promises that Genie 4 will be able to create interactive dialog that adapts as you change the simulation. Realistic speech in standard AI video was a major shift in realism, and now a similar revolution may be coming to Google’s world models.
Your very own world
DeepMind is careful to point out that world models are still an active area of research—Genie 3 never expanded beyond the very limited Project Genie web experience. However, the team has been working with a small group of trusted testers to build toward a wider release for Genie 4. Access will start expanding soon, but there’s no specific timeline yet. It won’t be cheap, though.
Google plans to roll out Genie 4 for developers via the Gemini API and AI Studio, but pricing isn’t yet available. Consumers will be able to get access, but it will roll out in phases with Google’s most expensive AI Ultra plans, which start at $100 per month. Google may expand access to AI Pro and AI Plus plans eventually. This will still be handled through Project Genie, which will limit how much you can do with Genie. There’s a waitlist for users who are interested in getting early access.

Kharkiv, Ukraine – Yulia Spichak, a 34 year-old makeup artist in Ukraine’s eastern Kharkiv region, dropped her son off at an English-language lesson on Oct. 19 and sat down in a nearby park to send a selfie to her sister, Alina. Seconds later, a Russian drone slammed into the park, killing Spichak and four others, including a 1-year-old girl.
The attack came six days after Elon Musk’s SpaceX throttled Ukraine’s access to Starlink, which Ukraine has been using to coordinate its defenses against Russian attacks and for some of its offensive operations. In a post on X, Musk accused Ukraine of violating Starlink’s terms of service by also using it for attacks on Russia.
Ukrainian officials told CBS News that the restrictions imposed without advance notice affected ground-based air defense radars and drone interceptors, hindering the military’s ability to stop incoming Russian strikes.
“It’s an essential part of the air defense infrastructure,” Vadym Ivchenko, a Ukrainian member of parliament on the National Security, Defense and Intelligence Committee, told CBS News Monday. “I couldn’t say that all other air defenses stopped working, no. But this part is quite essential for the defense of civilians in Ukraine.”
CBS News obtained photographs that show Ukraine’s ground-based air defense radars’ Starlink connectivity dropping to .1 megabytes per second on Oct. 19, the day of the Kharkiv drone attack. That would not be enough to load a webpage, let alone power air defense radars, according to a source familiar with the technology who requested anonymity because he was not permitted to comment on the matter.
On Saturday, Ukrainian officials confirmed that Starlink’s access had been restored across the country, with Musk admitting that SpaceX had throttled connectivity.
“Using Starlink, a civilian commercial service, for offensive drone strikes to kill people has never been allowed, nor should it be,” Musk wrote on X. “Ukraine military violated Starlink terms of service and used it for offensive drone strikes to kill people. That loophole was closed last week.”
Ukrainian officials have also complained of connectivity problems with Starshield, SpaceX’s military-grade network, which runs on the same constellation of satellites as Starlink and serves as the backbone of Ukraine’s drone operations. But Musk said the U.S. government controlled whatever changes happened with Starshield. SpaceX did not respond to a request for comment, and the White House referred CBS News to President Trump’s comments last week, during which he told reporters that he did not order the Starlink restrictions in Ukraine.
Alina Spichak, the sister of Yulia, the make-up artist who died in the strike in Kharkiv, revisited the park where she died with CBS News on Sunday. She said the Geran-4 Russian jet-powered drone that killed her sister arrived so quickly the victims didn’t hear it coming.
“I understand that everyone has their own lives and nobody wants to look at these horrors. But I believe that the whole world has to know what is happening in Ukraine, that there is a war here and we are being attacked,” Alina Spichak said. “I really hope that it will end soon.”

Apple’s upcoming iPhone 18e model will retain a 60Hz display but could gain an upgraded front-facing camera, according to the leaker known as “Digital Chat Station.”
In a new post on Weibo, the leaker said that the device will feature a 6.12-inch OLED display using low-temperature polycrystalline silicon (LTPS) technology, the A20 chip, and a single 48-megapixel rear camera. They were less definitive about the front-facing camera, saying it could be upgraded to 24 megapixels.
The specifications suggest that the iPhone 18e will offer no major display or rear camera hardware upgrades over the previous two generations. Both the iPhone 16e and iPhone 17e feature approximately 6.1-inch ‌OLED‌ displays with a 60Hz refresh rate and a single 48-megapixel rear camera. The post does not detail the rear camera’s sensor or lens, so smaller hardware changes remain possible.
The A20 chip would represent a more significant upgrade. Like the A20 Pro chip in the iPhone 18 Pro, the A20 is expected to use a 2nm manufacturing process, bringing notable performance and power efficiency improvements over the iPhone 17e’s 3nm A19 chip. It would also bring the device into line with the upcoming standard iPhone 18, which is expected to use the same chip.
The front-facing camera claim is less clear. Digital Chat Station specified 24 megapixels, although it is unclear whether this refers to a new camera or is a mistaken reference to the 18-megapixel Center Stage camera introduced with the iPhone 17 lineup. A 24-megapixel camera was widely rumored before those devices launched. Either resolution would represent an upgrade over the 12-megapixel front camera used by the iPhone 16e and ‌iPhone 17e‌.
Digital Chat Station previously claimed that the iPhone 18e would retain a 60Hz LTPS display. An earlier report from Korea suggested that an LTPO display capable of supporting 120Hz ProMotion could reach the entry-level iPhone with the “iPhone 19e” in 2028.
The iPhone 18e is expected to launch in spring 2027, alongside the standard ‌iPhone 18‌ and iPhone Air 2.
Related Roundup: iPhone 17e

Broadcom is providing “residual value support” to lenders, who are providing first lien loans or senior debt, effectively providing a significant guarantee in the financing structure.
Sign up here.
The latest deal highlights ​the growing role of banks and alternative asset managers in financing the infrastructure needed to ​support rising demand for AI computing capacity.
Anthropic filed for a potentially record-breaking IPO on ⁠Monday, even as disclosures of steep losses and security concerns threatened to test investor appetite for ​the AI heavyweight.
The size of the first lien loans or senior debt is about $43.5 billion, according to ​two sources familiar with the matter, or 69% of total financing.
Blackstone will provide the committed portion of the second-lien financing – which ranks below senior debt – while Bank of America, Citigroup and Morgan Stanley are arranging the first-lien financing, ​the statement said.
Chipmakers have provided guarantees in the past to absorb potential declines in the value of ​the assets, such as advanced chips and data infrastructure.
The release did not disclose the breakdown of the loans, while ‌Broadcom ⁠did not respond to a request for comment.
Rising AI demand is driving massive investment in computing infrastructure as technology companies race to secure capacity for increasingly complex workloads.
Bank of America, Citi and Morgan Stanley are arranging the first-lien financing.
Deliveries are expected to begin in 2027 as part of a previously announced ​strategic compute capacity agreement involving Broadcom, the companies said.
Apollo and Blackstone had ​agreed to finance ⁠a $35 billion expansion of Anthropic’s AI computing capacity under a prior agreement with Broadcom. Broadcom’s backstop in that deal was up to $29 billion, it said in a filing.
While Moody’s described Broadcom as one of the strongest ⁠large technology ​companies, it warned that the risk would increase if multiple ​transactions were undertaken over a short period to meet the sizable funding needs of frontier AI labs.
Reporting by Aditya Soni and ​Prakhar Srivastava in Bengaluru and Saeed Azhar in New York; Editing by Pooja Desai and Lincoln Feast.
Our Standards: The Thomson Reuters Trust Principles., opens new tab
Prakhar Srivastava covers U.S. finance, including banks, card companies, payment firms, insurers, exchanges and asset managers. He also writes on IPOs and venture capital. Previously, he worked on Reuters’ Global Pictures Desk and holds a degree in Political Science from the University of Delhi.
Saeed Azhar is a Reuters financial journalist and part of the U.S. banking team, which covers Wall Street’s biggest banks. He focuses on Goldman Sachs and Bank of America, and also writes about regional banks. Before moving to New York in July 2022, he led the finance team in the Middle East from Dubai, and also worked in Singapore, covering Southeast Asia finance.

Staffers across the Windows maker ‌would be eligible for the offers, including those working in AI experiences, Copilot, finance, human resources, Xbox and Windows + Devices, a source familiar with the matter told Reuters, requesting anonymity as the matter was ​private. Units such as marketing, commercial business and security will be excluded, the source ​added.
Sign up here.
“This offer gives eligible employees who may be considering a future outside ⁠the company a new choice to make that transition now with support. It also ​gives participating organizations another way to adapt their workforce to changing needs,” Chief People Officer Amy ​Coleman said in a staff memo.
Like other tech giants, Microsoft is spending hundreds of billions of dollars on AI data centers and chips, drawing scrutiny from investors over whether the returns will justify the staggering ​cost. The company, slated to report first-quarter results after markets close on Wednesday, has ​forecast $175 billion in capital expenditures for the 2026 calendar year.
Microsoft had 223,000 full-time employees globally as of June ‌30, ⁠2026, according to its annual filing.
Shares of Microsoft, which had risen about 10% this year, were up about 1% on Monday. The development was first reported by Business Insider.
The company earlier this year offered voluntary buyouts to about 7% of its US workforce, or about 9,000 employees.
In ​July, Microsoft announced it would ​cut 4,800 jobs, ⁠or about 2.1% of its global workforce, overhauling its Xbox gaming business and divesting up to five studios as it looks to boost ​returns after years of heavy investment in the division.
“This offer will ​be available ⁠to eligible employees in participating teams in Canada, India, the United Kingdom and the United States at level 70 and below,” Microsoft said on Monday.
The company said it was a time-limited, voluntary ⁠program ​and built on insights gained from a previous voluntary ​retirement program. The eligible employees will receive details within the next few hours, it added.
Reporting by Jaspreet Singh and ​Aditya Soni, additional reporting by Deborah Sophia in Bengaluru; Editing by Jonathan Ananda and Devika Syamnath