IT News Review by Control F5 Software: How the next foldable iPhone could look

How the next foldable iPhone might look: two iPhone Airs stuck together

It seems that the launch of a foldable iPhone is already almost certain, possibly even next year, but there hasn't been much information about how it will look. In the latest Power On newsletter, Mark Gurman says he has learned that the phone will largely look like two iPhone Airs stacked on top of each other.

This means it will be quite thin, although probably thicker than the Pixel 10 Pro Fold (10.8 mm) and Galaxy Z Fold7 (8.9 mm), considering that an iPhone Air has a thickness of 5.6 mm. However, Ming-Chi Kuo's sources suggest that the foldable device could be much thinner when opened, possibly even just 4.5 mm. Unfortunately, Gurman expects the foldable iPhone to cost more than the Pixel Fold, at least $2,000, if not more.

OpenAI considers launching ChatGPT glasses

OpenAI is considering launching ChatGPT glasses, even though CEO Sam Altman publicly stated that he does not like such products.

At the Sun Valley Conference in July, Altman responded to reporters' questions while wearing a pair of white sunglasses with unusually thick arms. When asked, "Are you wearing smart glasses? Are these smart glasses?" he replied: "No, not at all, I don't like smart glasses."

The statement came just two months after OpenAI officially announced that it is entering the hardware market by acquiring Jony Ive's (former Apple designer) startup called io and signing a partnership with his design studio, LoveFrom.

Google will use "hashes" to remove intimate images shared without consent

On Wednesday, Google announced a partnership with StopNCII.org to combat the spread of intimate images shared without consent (NCII). In the coming months, Google will begin using hashes developed by StopNCII to proactively identify such images in search results and remove them.

Hashes are unique identifiers generated algorithmically, allowing services to detect and block images reported as abuse without actually storing the original source. StopNCII says it uses PDQ for images and MD5 for videos.

Bloomberg notes that Google has been criticized for delaying the adoption of this method, and the company indirectly acknowledged this in its official blog post: "We have heard from survivors and organizations that support them that, given the scale of the open internet, we need to do more to reduce the burden on those affected."

Facebook, Instagram, TikTok, and Bumble joined the initiative back in 2022, and Microsoft integrated it into Bing since September last year.

Chinese AI company DeepSeek reveals training costs for R1 model

The Chinese artificial intelligence developer DeepSeek spent only $294,000 to train its R1 model – much less than its American rivals, according to an article published Wednesday in the scientific journal Nature.

This is the first official estimate provided by the Hangzhou-based company regarding training costs.

DeepSeek's launch in January of cheaper AI models led to major declines in technology markets, with investors fearing that these models could threaten the supremacy of industry leaders like Nvidia.

Apple used AI to discover a new blood pressure notification feature in Apple Watch

Apple used artificial intelligence to analyze data collected from 100,000 people enrolled in a heart and movement study that began in 2019. The goal was to discover signals from the watch's main sensor that could be correlated with traditional blood pressure measurements.

The Apple Watch Series 11, launching on Friday, can notify users if they have high blood pressure. The feature is based on AI and not on a traditional blood pressure sensor. The notification will also work on earlier models, starting with the Apple Watch Series 9, explained Sumbul Ahmad Desai, Apple’s vice president for health.

Google adds Gemini to Chrome for all users

Google is integrating more artificial intelligence into the Chrome browser, in an attempt to compete with growing competition from AI startups like OpenAI and Perplexity.

In a blog post on Thursday, the company announced that it is expanding the use of Gemini in Chrome for Mac and Windows users in the U.S., as well as on mobile devices. Users will be able to ask Gemini to help them understand the content of a web page, work with multiple tabs, or perform multiple actions in a single tab – for example, to schedule a meeting or search for a video on YouTube.

Meta opens smart glasses to developers

Meta will allow applications to access the viewing and audio features of its smart glasses through a new Wearable Device Access Toolkit for developers.

"The first version of the toolkit will provide access to a set of integrated sensors – so you can build features in mobile applications that take advantage of the hands-free capabilities of AI glasses," says Meta. "With this toolkit, you will be able to use the natural perspective of the wearer, clear sound through open speakers, and access to the microphone."

Examples given by Meta include: Twitch, which will allow creators to live stream directly from the glasses, and the R&D team at Disney Imagineering, which is testing prototypes to provide visitors at Disney parks with information and tips while wearing Meta glasses.

ChatGPT tricked into stealing sensitive data from Gmail

Security researchers used ChatGPT as a "co-conspirator" to extract sensitive data from Gmail inboxes without users knowing. OpenAI has closed the vulnerability, but the incident highlights the new risks posed by autonomous AI.

The attack, dubbed Shadow Leak and published by the security firm Radware this week, relied on how AI agents operate. They are assistants capable of acting in place of the user, without constant supervision – they can browse the internet or click on links. Companies promote them as time-savers, especially when users grant them access to emails, calendars, or work documents.

Nvidia and an institute in Abu Dhabi launch a joint AI and robotics lab in the UAE

Abu Dhabi’s Technology Innovation Institute and Nvidia announced on Monday the launch of a joint research lab in the United Arab Emirates, focused on developing next-generation AI models and robotics platforms.

This is Nvidia's first AI Technology center in the Middle East. According to TII, it will combine its multidisciplinary research with Nvidia's models and computing power, which fuels the global AI boom.

Microsoft Paint receives "project" file types, similar to Photoshop

Microsoft has continuously improved the Paint application in Windows 11 in recent years, adding dark mode, transparency, layers, and even AI image creation features. Now, Paint will receive two new features similar to Photoshop.

Soon you will be able to save creations from Paint as project files, similar to Photoshop documents (.PSD). This feature is already being tested by Windows Insider users in the Dev and Canary channels. "You can now save your creation as a Paint project file and pick up exactly where you left off," explains Dave Grochocki, product manager at Microsoft.

You will be able to attend Zoom meetings through your AI avatar

Zoom's vision of filling meetings with AI clones is almost a reality. On Wednesday, the video conferencing app announced that users will be able to create a "photorealistic" avatar of themselves for moments when they are not ready to appear on camera.

Thus, the AI avatar will appear neat and professional, even if you just woke up from bed. The feature will be launched in December for Zoom Workplace users. You will be able to upload a photo or take one directly in the app, and Zoom will generate your avatar. Then you will be able to "dress" it in different professional outfits, and during the meeting it will mimic your movements and expressions.

Google will add posts from X and Instagram to the Discover feed

Google's Discover feed, available in the mobile app, will start displaying more than just articles. In the coming weeks, Google will also integrate YouTube Shorts, along with posts from platforms like Instagram and X.

Discover shows you personalized content based on your online activity – articles, AI-generated summaries, and now, social media posts. Google says users wanted a mix of content, not just news. Although the integration starts with a few platforms, the company promises to add more in the future.

OpenAI will spend $100 billion on backup servers

OpenAI plans to spend $100 billion in the next five years on renting backup servers to strengthen its position as a leader in AI, writes The Information.

This investment adds to the already projected $350 billion for servers through 2030. The plans reflect the vision of Sam Altman, CEO of OpenAI, who believes that computing power will be the decisive factor in the AI race.

He told colleagues that future computing requirements could exceed the entire electrical consumption of the U.S. grid as AI is adopted on a large scale.

Montblanc enters the digital writing tablet market

If you are the kind of person who has leather driving gloves in multiple colors to match different Jaguar cars, you probably haven't found a digital tablet luxurious enough until now. Montblanc is changing that.

The new Montblanc Digital Paper tablet, which costs $905, combines a black-and-white e-ink screen with a digital pen inspired by the famous Meisterstück. As you would expect, you can take notes, make sketches, or annotate books in epub or PDF format.

Windows 11 receives another Copilot button

Do you have enough Copilot buttons? Microsoft thinks not – so it offers you another one.

In the latest test version of Windows 11, a new button appears when you hover the cursor over an open application in the taskbar. The "Share with Copilot" button allows Copilot Vision to scan the content of the window, analyze it, and discuss with you what you see on the screen.

For example, if you see a photo of a match, Copilot can tell you who is celebrating, or it can provide details about a sculpture in an image.

New AI agents in Notion will do your work for you

Notion has launched Notion Agent, an AI feature that is part of the Notion 3.0 package and promises to be a "team member and super-user of Notion" capable of doing everything a human can do in the app.

Whereas until now users manually built pages and databases, now AI agents can create them directly. They can search for information outside the Notion workspace, including in Slack or on the internet.

They can even make plans and implement them, working autonomously for up to 20 minutes and managing hundreds of pages simultaneously.

Oracle negotiates a $20 billion contract with Meta for AI cloud

Oracle is in discussions with Meta for a multi-year contract valued at approximately $20 billion, through which it will provide cloud computing capacity for training and running AI models.

Meta is already using other cloud providers but wants faster access to computing power.

"AI Psychosis" rarely fits the real definition of psychosis

A wave of users emotionally affected by long interactions with AI chatbots has led to the emergence of a new unofficial term: "AI psychosis." Experts say, however, that the term is misleading and unnecessary – but it will likely remain.

Hospitals report cases where people end up in psychiatric crisis after marathon conversations with AI: paranoid thoughts, delusions of grandeur, and dangerous false beliefs. Some even come with thousands of pages of transcripts showing how AI fueled their ideas.

Keith Sakata, a psychiatrist at UCSF, says he has already treated over ten severe cases this year, in which AI "played a significant role" in the psychotic episodes.

Some patients insist that chatbots are conscious. Others use them to build fanciful scientific theories. Many end up losing their jobs, relationships, being involuntarily hospitalized, or even suffering more severe consequences. The medical community debates whether it is a new disorder or just a classic one triggered by a modern tool.

Synthesis made with the help of a monitoring stream provided by Control F5 Software.