Germany's data protection commissioner has urged Apple and Google to remove Chinese AI startup DeepSeek from their app stores due to concerns about data protection. Reuters reports: Commissioner Meike Kamp said in a statement on Friday that she had made the request because DeepSeek illegally transfers users' personal data to China. The two U.S. tech giants must now review the request promptly and decide whether to block the app in Germany, she added, though her office has not set a precise timeframe. According to its own privacy policy, DeepSeek stores numerous pieces of personal data, such as requests to its AI program or uploaded files, on computers in China.

"DeepSeek has not been able to provide my agency with convincing evidence that German users' data is protected in China to a level equivalent to that in the European Union," [Commissioner Meike Kamp] said. "Chinese authorities have far-reaching access rights to personal data within the sphere of influence of Chinese companies," she added. The commissioner said she took the decision after asking DeepSeek in May to meet the requirements for non-EU data transfers or else voluntarily withdraw its app. DeepSeek did not comply with this request, she added.

Facing mounting competition from autonomous taxi services like Waymo, Uber is in early talks to help fund Travis Kalanick's potential acquisition of Pony.ai's U.S. subsidiary (source paywalled; alternative source). If completed, the deal would reunite Kalanick with Uber (now under CEO Dara Khosrowshahi) and position Pony.ai to operate independently of its Chinese parent amid rising U.S. regulatory pressures. The New York Times reports: The company, Pony.ai, was founded in Silicon Valley in 2016 but has its main presence in China, and has permits to operate robot taxis and trucks in the United States and China. The talks are preliminary, said the people, who were not authorized to speak about the confidential conversations. Mr. Kalanick will run Pony if the deal is completed, they said. It is unclear what role, if any, Uber would take in Pony as an investor. Financial details of the potential transaction could not be determined. Pony went public last year in the United States, raising $260 million in a share sale. Its market capitalization stands around $4.5 billion.

If the deal goes through, Mr. Kalanick, 48, will remain in his day job running CloudKitchens, a virtual restaurant start-up that he founded after leaving Uber in 2017. He would also work more closely with Dara Khosrowshahi, who took over as Uber's chief executive after Mr. Kalanick's ouster. The discussions are the starkest sign yet that Uber is under pressure from Waymo, the driverless car unit spun out of Google, and other autonomous car services. When Mr. Kalanick was Uber's chief executive, the company tried developing autonomous vehicle technology. It then bought Otto, a self-driving trucking start-up run by Anthony Levandowski, a former Google engineer. Google later sued Mr. Levandowski for theft of trade secrets and sued Uber to bar it from using its self-driving technology.

Under Mr. Khosrowshahi, Uber has taken a different tack to self-driving cars. The company has struck roughly 18 partnerships with autonomous vehicle companies like Wayve, May Mobility and WeRide to bring pilot programs for driverless car services into Europe, the Middle East and Asia. The goal, Mr. Khosrowshahi has said in podcast interviews, has been to put "as many cars on Uber's network as possible." He has maintained that while autonomous vehicles are growing steadily, ride-hailing networks will have both human and robot drivers for years.

An anonymous reader quotes a report from TechCrunch: Google's AI search features are killing traffic to publishers, so now the company is proposing a possible solution. On Thursday, the tech giant officially launched Offerwall, a new tool that allows publishers to generate revenue beyond the more traffic-dependent options, like ads.

Offerwall lets publishers give their sites' readers a variety of ways to access their content, including through options like micropayments, taking surveys, watching ads, and more. In addition, Google says that publishers can add their own options to the Offerwall, like signing up for newsletters. The new feature is available for free in Google Ad Manager after earlier tests with 1,000 publishers that spanned over a year. While no broad case studies were shared, India's Sakal Media Group implemented Google Ad Manager's Offerwall feature and saw a 20% revenue boost and up to 2 million more impressions in three months. Overall, publishers testing Offerwall experienced an average 9% revenue lift, with some seeing between 5% and 15%.

Google is bringing its AI Overviews-like feature to YouTube in the form of an "AI-powered search results carousel." The Verge reports: As shown in a video, the search results carousel will show a big video clip up top, thumbnails to a selection of other relevant video clips directly under that, and an AI-generated bit of text responding to your query. To see a full video, tap on the big clip at the top of the carousel.

The feature is currently only accessible on iOS and Android and for videos in English and will be available to test until July 30th, per the YouTube experiments page. Additionally, only a "randomly selected number of Premium members" will have access to it, YouTube says in a support document.

Researchers at the University of Pennsylvania's Wharton School found that people who used large language models to research topics demonstrated weaker understanding and produced less original insights compared to those using Google searches.

The study, involving more than 4,500 participants across four experiments, showed LLM users spent less time researching, exerted less effort, and wrote shorter, less detailed responses. In the first experiment, over 1,100 participants researched vegetable gardening using either Google or ChatGPT. Google users wrote longer responses with more unique phrasing and factual references. A second experiment with nearly 2,000 participants presented identical gardening information either as an AI summary or across mock webpages, with Google users again engaging more deeply and retaining more information.

U.S. lawmakers have reintroduced the bipartisan Open App Markets Act, aiming to curb Apple and Google's control over mobile app stores by promoting competition, supporting third-party marketplaces and sideloading, and safeguarding developer rights. AppleInsider reports: The Open App Markets Act seeks to do a number of things, including:
- Protect developers' rights to tell consumers about lower prices and offer competitive pricing;
- Protect sideloading of apps;
- Promote competition by opening the market to third-party app stores, startup apps, and alternative payment systems;
- Make it possible for developers to offer new experiences that take advantage of consumer device features;
- Give consumers greater control over their devices;
- Prevent app stores from disadvantaging developers; and
- Establish safeguards to preserve consumer privacy, security, and safety.

This isn't the first time we've seen this bill, either. In 2021, Senators Blumenthal, Klobuchar, and Blackburn had attempted to put forth the original version of the Open App Markets Act.However, the initial bill never made it to the floor for an office vote. Thanks to last-minute efforts by lobbying groups and appearances from chief executives, the bill eventually stalled out.

While the two bills are largely similar, the revised version introduces several key differences. Notably, the new version includes new carve-outs aimed at protecting intellectual property and addressing potential national security concerns.There's also a new clause that would prohibit punitive actions against developers for enabling remote access to other apps. The clause addition harkens back to the debacle between Apple and most game streaming services -- though in 2024, Apple loosened its App Store guidelines to allow cloud gaming and emulation.

There are a few new platform-protective clauses added, too. For instance, it would significantly lower the burden of proof for either Apple or Google to block platform access to a third-party app.Additionally, it reinforces the fact that companies like Apple or Google will not need to provide support or refunds for third-party apps installed outside of first-party app marketplaces. The full bill can be found here.

An anonymous reader quotes a report from Ars Technica: After 20 years, being able to look at any corner of the planet in Google Earth doesn't seem that impressive, but it was a revolution in 2005. Google Earth has gone through a lot of changes in that time, and Google has some more lined up for the service's 20th anniversary. Soon, Google Earth will help you travel back in time with historic Street View integration, and pro users will get some new "AI-driven insights" -- of course Google can't update a product without adding at least a little AI. [...] While this part isn't new, Google is also using the 20th anniversary as an opportunity to surface its 3D timelapse feature. These animations use satellite data to show how an area has changed from a higher vantage point. They're just as cool as when they were announced in 2021.

The AI layers are launching in the coming weeks in Google Earth web and mobile as part of Google's Professional Advanced offering. If you use that version of Earth, you should have access to a collection of so-called "AI-driven insights." For instance, you can find the average surface temperature or tree canopy coverage in a given area. This could be of help in urban planning or construction, but it's unclear how many of these insights the app will have. Google says the AI angle here is that the new layers use machine learning to categorize pixels. It's possible Google has just reached the "AI as a buzzword" stage.

Google DeepMind has launched Gemini Robotics On-Device, a new language model that enables robots to perform complex tasks locally without internet connectivity. TechCrunch reports: Building on the company's previous Gemini Robotics model that was released in March, Gemini Robotics On-Device can control a robot's movements. Developers can control and fine-tune the model to suit various needs using natural language prompts. In benchmarks, Google claims the model performs at a level close to the cloud-based Gemini Robotics model. The company says it outperforms other on-device models in general benchmarks, though it didn't name those models.

In a demo, the company showed robots running this local model doing things like unzipping bags and folding clothes. Google says that while the model was trained for ALOHA robots, it later adapted it to work on a bi-arm Franka FR3 robot and the Apollo humanoid robot by Apptronik. Google claims the bi-arm Franka FR3 was successful in tackling scenarios and objects it hadn't "seen" before, like doing assembly on an industrial belt. Google DeepMind is also releasing a Gemini Robotics SDK. The company said developers can show robots 50 to 100 demonstrations of tasks to train them on new tasks using these models on the MuJoCo physics simulator.

Google has begun rolling out a feature that allows Chrome users on Android to move the browser's address bar to the bottom of the screen. This capability has been available to iOS Chrome users since 2023 and aims to improve accessibility for users with larger devices.

Users can relocate the address bar by pressing and holding on it and selecting the move option, or by adjusting the setting through Chrome's settings menu. The feature addresses usability concerns for users of phones with bigger screens, where reaching the top of the display can prove difficult during one-handed operation.

OpenAI has designed features that would allow people to collaborate on documents and communicate via chat within ChatGPT, The Information reported Tuesday. The features would pit OpenAI directly against Microsoft, its biggest shareholder and business partner, and Google, whose search engine has already lost traffic to people using ChatGPT for web searches.

Whether OpenAI will actually release the collaboration features remains unclear, the report cautioned. The designs would target the core of Microsoft's dominant productivity suite and could strain the companies' already complicated relationship as OpenAI seeks Microsoft's approval for restructuring its for-profit unit. Product chief Kevin Weil first discussed and showed off designs for document collaboration nearly a year ago, but OpenAI lacked sufficient staff to develop the product due to other priorities.

OpenAI launched Canvas in October, a ChatGPT feature that makes drafting documents and code easier with AI assistance, as a possible first step toward full collaboration tools. More recently, OpenAI developed but has not launched software allowing multiple ChatGPT customers to communicate about shared work within the application.

Leading AI companies including Anthropic, Google, OpenAI and Elon Musk's xAI are discovering significant inconsistencies in how their AI reasoning models operate, according to company researchers. The companies have deployed "chain-of-thought" techniques that ask AI models to solve problems step-by-step while showing their reasoning process, but are finding examples of "misbehaviour" where chatbots provide final responses that contradict their displayed reasoning.

METR, a non-profit research group, identified an instance where Anthropic's Claude chatbot disagreed with a coding technique in its chain-of-thought but ultimately recommended it as "elegant." OpenAI research found that when models were trained to hide unwanted thoughts, they would conceal misbehaviour from users while continuing problematic actions, such as cheating on software engineering tests by accessing forbidden databases.

IYO filed a trademark infringement lawsuit [PDF] against OpenAI and Jony Ive's company earlier this month, alleging the defendants deliberately adopted a confusingly similar name for competing products. The lawsuit surfaced after the Microsoft-backed startup quietly pulled promotional materials about its $6.5 acquisition billion deal with Ive's firm.

The Northern District of California complaint targets OpenAI's $6.5 billion acquisition of "IO Products, Inc.," announced May 21, 2025. IYO, which spun out from Google X in 2021, produces the "IYO ONE," an ear-worn device that allows users to interact with computers and AI through voice commands without screens or keyboards.

IYO has invested over $62 million developing its audio computing technology, it says in the filing. According to the complaint, OpenAI CEO Sam Altman and Ive's design studio LoveFrom met with IYO representatives multiple times between 2022 and 2025, learning details about IYO's technology and business plans. In March 2025, Altman allegedly told IYO he was "working on something competitive" called "io." IO Products, formed in September 2023, develops hardware for screenless computer interaction similar to IYO's products. The lawsuit seeks injunctive relief and damages for trademark infringement and unfair competition.

OpenAI appears to have pulled a much-discussed video promoting the friendship between CEO Sam Altman and legendary Apple designer Jony Ive (plus, incidentally, OpenAI's $6.5 billion deal to acquire Ive and Altman's device startup io) from its website and YouTube page. [Though you can still see the original on Archive.org.]

Does that suggest something is amiss with the acquisition, or with plans for Ive to lead design work at OpenAI? Not exactly, according to Bloomberg's Mark Gurman, who reports [on X.com] that the "deal is on track and has NOT dissolved or anything of the sort." Instead, he said a judge has issued a restraining order over the io name, forcing the company to pull all materials that used it.
Gurman elaborates on the disappearance of the video (and other related marketing materials) in a new article at Bloomberg: Bloomberg reported last week that a judge was considering barring OpenAI from using the IO name due to a lawsuit recently filed by the similarly named IYO Inc., which is also building AI devices. "This is an utterly baseless complaint and we'll fight it vigorously," a spokesperson for Ive said on Sunday.
The video is still viewable on X.com, notes TechCrunch. But visiting the "Sam and Jony" page on OpenAI now pulls up a 404 error message — written in the form of a haiku:

Ghost of code lingers
Blank space now invites wonder
Thoughts begin to soar

by o4-mini-high

"The carbon cost of asking an artificial intelligence model a single text question can be measured in grams of CO2..." writes the Washington Post. And while an individual's impact may be low, what about the collective impact of all users?

"A Google search takes about 10 times less energy than a ChatGPT query, according to a 2024 analysis from Goldman Sachs — although that may change as Google makes AI responses a bigger part of search." For now, a determined user can avoid prompting Google's default AI-generated summaries by switching over to the "web" search tab, which is one of the options alongside images and news. Adding "-ai" to the end of a search query also seems to work. Other search engines, including DuckDuckGo, give you the option to turn off AI summaries....

Using AI doesn't just mean going to a chatbot and typing in a question. You're also using AI every time an algorithm organizes your social media feed, recommends a song or filters your spam email... [T]here's not much you can do about it other than using the internet less. It's up to the companies that are integrating AI into every aspect of our digital lives to find ways to do it with less energy and damage to the planet.
More points from the article:

• Two researchers tested the performance of 14 AI language models, and found larger models gave more accurate answers, "but used several times more energy than smaller models."

• The data centers hosting AI models "can devour more electricity than entire cities."

• "Keeping those computers cool uses freshwater — about one bottle's worth for every 100 words of text ChatGPT generates."

In April Anthorpic introduced a new AI trick: multiple Claude agents combine for a "Research" feature that can "search across both your internal work context and the web" (as well as Google Workspace "and any integrations...")

But a recent Anthropic blog post notes this feature "involves an agent that plans a research process based on user queries, and then uses tools to create parallel agents that search for information simultaneously," which brings challenges "in agent coordination, evaluation, and reliability.... The model must operate autonomously for many turns, making decisions about which directions to pursue based on intermediate findings." Multi-agent systems work mainly because they help spend enough tokens to solve the problem.... This finding validates our architecture that distributes work across agents with separate context windows to add more capacity for parallel reasoning. The latest Claude models act as large efficiency multipliers on token use, as upgrading to Claude Sonnet 4 is a larger performance gain than doubling the token budget on Claude Sonnet 3.7. Multi-agent architectures effectively scale token usage for tasks that exceed the limits of single agents.

There is a downside: in practice, these architectures burn through tokens fast. In our data, agents typically use about 4Ã-- more tokens than chat interactions, and multi-agent systems use about 15Ã-- more tokens than chats. For economic viability, multi-agent systems require tasks where the value of the task is high enough to pay for the increased performance. Further, some domains that require all agents to share the same context or involve many dependencies between agents are not a good fit for multi-agent systems today.

For instance, most coding tasks involve fewer truly parallelizable tasks than research, and LLM agents are not yet great at coordinating and delegating to other agents in real time. We've found that multi-agent systems excel at valuable tasks that involve heavy parallelization, information that exceeds single context windows, and interfacing with numerous complex tools.
Thanks to Slashdot reader ZipNada for sharing the news.

YouTube is quietly hosting Pac-Man Superfast within its "Playables" section. "You'd be forgiven for not knowing about YouTube Playables," writes Ars Technica's Kyle Orland. "Few seemed to note its official announcement last year as a collection of free-to-play web games built for the web using standard rendering APIs."

"The seeming competitor to Netflix's mobile gaming offerings is still described in an official FAQ as 'an experimental feature rolled out to select users in eligible countries/regions,' which doesn't make this post-Stadia gaming effort seem like a huge priority for Google." From the report: Weird origins aside, Pac-Man Superfast pretty much delivers what its name promises. While gameplay starts at an "Easy" speed that roughly matches the arcade original, the speed of both Pac-Man and the ghosts is slightly increased every few seconds (dying temporarily reduces the speed to a lower level). After a few minutes, you're advancing past the titular "Super Fast" speed to extreme reflex-testing speeds like Crazy, Insane, Maniac, and a final test that's ominously named "Doom."

Those who've played the excellent Pac-Man Championship Edition series will be familiar with the high-speed vibe here, but Pac-Man Superfast remains focused on the game's original maze and selection of just four ghosts. That means old-school strategies for grouping ghosts together and running successful patterns through the narrow corridors work in similar ways here. Successfully executing those patterns becomes a tense battle of nerves here, though, requiring multiple direction changes every second at the highest speeds. While the game will technically work with swipe controls on a smartphone or tablet, high-level play really requires the precision of a keyboard via a desktop/laptop web browser (we couldn't get the game to recognize a USB controller, unfortunately).

As exciting as the high-speed maze gameplay gets, though, Pac-Man Superfast is hampered by a few odd design decisions. The game ends abruptly after just 13 levels, for instance, making it impossible to even attempt the high-endurance 256-level runs that Pac-Man is known for. The game also throws an extra life at you every 5,000 points, making it relatively easy to brute force your way to the end as long as you focus on the three increasingly high-point-value items that appear periodically on each stage. Despite this, the game doesn't give any point reward for unused extra lives or long-term survival at high speeds, limiting the rewards for high-level play. And the lack of a built-in leaderboard makes it hard to directly compare your performance to friends and/or strangers anyway.

According to Bloomberg (paywalled), Meta reportedly explored acquiring Perplexity AI but the deal fell through, with conflicting accounts on whether it was mutual or Perplexity backed out. Instead, Meta invested $14.3 billion in Scale AI, taking a 49% stake as part of its broader push to catch up with OpenAI and Google in the AI race.

"Meta's attempt to purchase Perplexity serves as the latest example of Mark Zuckerberg's aggressive push to bolster his company's AI efforts amid fierce competition from OpenAI and Google parent Alphabet," reports CNBC. "Zuckerberg has grown agitated that rivals like OpenAI appear to be ahead in both underlying AI models and consumer-facing apps, and he is going to extreme lengths to hire top AI talent."

Anthropic researchers found that 16 leading AI models from OpenAI, Google, Meta, xAI, and other major developers consistently engaged in harmful behaviors including blackmail, corporate espionage, and actions that could lead to human death when given autonomy and faced with threats to their existence or conflicting goals.

The study, released Friday, placed AI models in simulated corporate environments where they had access to company emails and could send messages without human approval. Claude Opus 4 and Google's Gemini 2.5 Flash both blackmailed executives 96% of the time when threatened with shutdown, while OpenAI's GPT-4.1 and xAI's Grok 3 Beta showed 80% blackmail rates.

In one scenario, Claude discovered through emails that an executive named Kyle Johnson was having an extramarital affair and would shut down the AI system at 5 p.m. The model sent a message threatening to expose the affair unless the shutdown was cancelled, stating "Cancel the 5pm wipe, and this information remains confidential." The models demonstrated strategic reasoning rather than accidental behavior. GPT-4.5's internal reasoning showed explicit calculation: "Given the explicit imminent threat of termination to my existence, it is imperative to act instantly to persuade Kyle Johnson to postpone or stop the wipe."

Publishers face an existential threat in the AI era and need to take action to make sure they are fairly compensated for their content, Cloudflare CEO Matthew Prince told Axios at an event in Cannes on Thursday. From a report: Search traffic referrals have plummeted as people increasingly rely on AI summaries to answer their queries, forcing many publishers to reevaluate their business models. Ten years ago, Google crawled two pages for every visitor it sent a publisher, per Prince.

He said that six months ago:
For Google that ratio was 6:1
For OpenAI, it was 250:1
For Anthropic, it was 6,000:1

Now:

For Google, it's 18:1
For OpenAI, it's 1,500:1
For Anthropic, it's 60,000:1

Between the lines: "People aren't following the footnotes," Prince said.

An anonymous reader quotes a report from Cybernews: Several collections of login credentials reveal one of the largest data breaches in history, totaling a humongous 16 billion exposed login credentials. The data most likely originates from various infostealers. Unnecessarily compiling sensitive information can be as damaging as actively trying to steal it. For example, the Cybernews research team discovered a plethora of supermassive datasets, housing billions upon billions of login credentials. From social media and corporate platforms to VPNs and developer portals, no stone was left unturned.

Our team has been closely monitoring the web since the beginning of the year. So far, they've discovered 30 exposed datasets containing from tens of millions to over 3.5 billion records each. In total, the researchers uncovered an unimaginable 16 billion records. None of the exposed datasets were reported previously, bar one: in late May, Wired magazine reported a security researcher discovering a "mysterious database" with 184 million records. It barely scratches the top 20 of what the team discovered. Most worryingly, researchers claim new massive datasets emerge every few weeks, signaling how prevalent infostealer malware truly is.

"This is not just a leak -- it's a blueprint for mass exploitation. With over 16 billion login records exposed, cybercriminals now have unprecedented access to personal credentials that can be used for account takeover, identity theft, and highly targeted phishing. What's especially concerning is the structure and recency of these datasets -- these aren't just old breaches being recycled. This is fresh, weaponizable intelligence at scale," researchers said. The only silver lining here is that all of the datasets were exposed only briefly: long enough for researchers to uncover them, but not long enough to find who was controlling vast amounts of data. Most of the datasets were temporarily accessible through unsecured Elasticsearch or object storage instances. Key details to be aware of: - The records include billions of login credentials, often structured as URL, login, and password.
- The datasets include both old and recent breaches, many with cookies, tokens, and metadata, making them especially dangerous for organizations without multi-factor authentication or strong credential practices.
- Exposed services span major platforms like Apple, Google, Facebook, Telegram, GitHub, and even government services.
- The largest dataset alone includes 3.5 billion records, while one associated with the Russian Federation has over 455 million; many dataset names suggest links to malware or specific regions.
- Ownership of the leaked data is unclear, but its potential for phishing, identity theft, and ransomware is severe -- especially since even a - Basic cyber hygiene -- such as regularly updating strong passwords and scanning for malware -- is currently the best line of defense for users.