Latest AI News
OpenAI’s first hardware device is reportedly a screenless speaker that can move
OpenAI’s first foray into hardware devices is reported to be a mobile smart speaker with integrated AI capabilities that can sync with ChatGPT and provide other home AI services. BloombergreportedTuesday that the device — which is still currently under development — is designed to be screen-free and is being pitched internally as a “humanlike AI companion that lives in the home.” OpenAI has long claimed that it wants to launch a hardware product — with some rumors being that it wants to launch its own phone, a move that would put it in competition with Apple. OpenAI’s newly surfaced device sounds like something of a departure from traditional smart speakers — as sources described the device to Bloomberg as having a “personality” and being able to proactively learn about its owner over time, providing more personalized service. The machine would have access to a user’s digital life, drawing off things like emails, sources said. The device is also weirdly described as involving “mechanical elements that can move on their own” and the Bloomberg report includes the detail that the device is designed to “feel like a companion and become a physical manifestation of OpenAI’s ChatGPT.” The device was developed with help from many former Apple engineers who were instrumental in “creating products such as the iPhone and Mac,” Bloomberg writes. Indeed, OpenAI may be attempting to launch a new hardware line, but the company is currently up to its eyeballs in trouble over hardware-related legal problems. Applelast week sued OpenAI, accusing the AI company of stealing its trade secrets. Apple further claimed that the allegations involved in the suit are merely“the tip of the iceberg” and that more misconduct will be revealed during the legal discovery process. OpenAI has denied wrongdoing. Citing anonymous sources with knowledge of OpenAI’s plans, Bloomberg writes that the company feels its new product “veers significantly from anything Apple has on the market today” and that it is “unlikely that it violates trade secrets” belonging to Apple. OpenAI’s push comes as the tech world grows more excited about consumer AI hardware more broadly. Hark, an AI lab founded by Brett Adcock, raised an oversubscribed $700 million Series A back in May at a$6 billion valuationto build what it calls “personal intelligence” — proprietary AI models paired with custom hardware designed as a “universal interface between humans and machines.” The company hasn’t yet detailed its device’s form factor, underscoring how much capital is chasing this category even before products ship.
View
New York State halts construction of all new data centers
New York became the first state to halt data center construction after Gov. Kathy Hochul signed an executive order today that temporarily bars the state from approving new permits for large projects. Hochul’s order applies to data centers 50 megawatts or larger, potentially affecting more than a dozen projects. The state’s Department of Environmental Conservation will not issue any permits that haven’t already been completed. While resource concerns have fueled some of the backlash, broader concern about AI has been behind much of it as well. A recentPew Research reportfound that only 10% of Americans were more excited than concerned about AI use in daily life, and just 23% felt that the technology would have a positive impact on how people do their jobs. Less than a quarter of the general public feels that AI will give the economy a boost, and less than a third were confident that the government would regulate the technology responsibly. “Progress shouldn’t arrive with a higher utility bill, deleted water supply, or noise pollution,” Hochul said at a press conference in Brooklyn. “These data centers can only be built, should only be built in places that want them. So they will never be exempt from local zoning, local approvals.” The moratorium will be lifted once the state finalizes an environmental review process for data centers, which Hochul expects will take about a year. Hochul’s office is also considering requiring data centers to pay into a fund that would support the state’s electrical grid, and she would like to prevent hyperscale data centers from receiving tax benefits. Hochul’s executive order arrives as more stringent measures are moving through New York’s legislature. Last month, the legislature advanced a bill that would pause construction of data centers larger than 20 megawatts for one year, while another still in committee would institute a three-year moratorium. The average data center built in the last few years has been smaller than 100 megawatts, but those in development are expected to be much larger as AI drives computing demands higher. Through 2030, nearly a quarter of new data centerswill exceed 500 megawatts, according to BloombergNEF, driven by increasing AI investment. The idea of a data center moratorium has been debated at the state and federal levels, but New York is the first to put one into practice. In December, more than 230 organizations called for a nationwide pause on new data centers. Vermont Sen. Bernie Sanders has also proposed a nationwide moratorium, though it hasn’t received much traction. More recently, Maine’s legislature passed a bill that would have paused construction on new data centers until November 1, 2027, but Gov. Janet Mills vetoed it. Just years ago, data centers were sought after by states eager to secure new development projects, but recently,public sentiment on data centers has souredas new projects have grown in size. The scale and pace at which they’re being constructed has started tostrain the electrical gridin addition to regional resources like water and farmland. Two-thirds of respondents to a recent poll said they wereconcerned about data centers driving up electricity prices. Another survey found that people would rather have an Amazon warehouse in their backyardthan a data center. Hochul’s order could be setting up for a clash with the Trump administration, which thus far has supported data center development. Last month, the Federal Energy Regulatory Commission, which is led by a Trump appointee, told grid operators todevelop special fast lanesto speed data centers’ interconnections.
View
Google Images gets a Pinterest-like redesign focused on discovery
Google Images, the tech giant’s image search engine, is taking on Pinterest with its latest redesign that turns the site into a browsable, dynamic gallery of images from across the web. Google is also adding a way for users to create AI images right in Search, as it celebrates 25 years since the debut of Google Images. Pinterest has long been known for allowing people to browse and save visual inspiration for everything from fashion to home decor. With this redesign, Google is essentially copying that playbook by turning Google Images into a place for discovery and inspiration, and not just search, which could increase users’ time spent on Google platforms, helping boost its ad revenue. In addition, Google is likely hoping that when users can’t find the image they’re looking for on Google Images or when they want to visualize something, they’ll stay within its ecosystem to create it rather than turn to third-party services like ChatGPT. After navigating to the redesigned Google Images, users will see a “For You” gallery of images tailored to their interests and browsing history. Like Pinterest, the gallery is designed for continuous browsing, with Google saying it updates in real time with new images. As users browse, they can save ideas to their “collections,” which will appear as tabs above the main gallery of photos. For example, users can create collections for things like vacation outfit ideas, travel inspiration, and ways to design a reading nook, which they can come back to later. The redesign is rolling out over the coming weeks on desktop in the U.S. in English. Users need to be signed into a Google Account to try it out, the tech giant says. As for generating images directly in Search, Google says the feature is meant for moments when you have a highly specific idea for an image that doesn’t already exist online. Google is bringing image generation directly into AI Overviews on Search and will use its latest Nano Banana model to transform a text prompt into a custom visual. The feature can also help users reimagine spaces and visualize ideas, such as seeing what a room might look like painted red or what a dorm room with a coastal theme could look like. Image generation in AI Overviews will start to roll out over the coming weeks in English for all regions that currently support image creation in AI Mode, Google says.
View
Meta’s Adam Mosseri says AI token budgets could soon be capped per engineer
In a recent interview, Instagram head Adam Mosseri said he can see a time in the future, perhaps only a year or two, when putting limits on Meta employees’ AI token spend will become necessary. “I think that you can imagine, at least in a year or two … that the burn rate of a strong engineer might be the same as their salary, or their cost of employment. And in that world, you’re going to probably need to put in some caps,” the Meta executive said, whilespeakingon Lenny’s Podcast. AI token spend, a reference to the cost of processing AI prompts and responses, has beena much-buzzed-about subjectin recent days. Meta shut downan internal AI token spend leaderboardafter AI costs put the company on track forbillionsof dollars in 2026. Meta is not alone in rethinking its approach to AI experimentation. Uber alsohad an AI reckoningafter it blew through its 2026 AI coding budget by April. Soaring token costs sawMicrosoft cancelClaude Code licenses, consolidating its engineers around its own Copilot CLI tool instead. Mosseri’s belief, he explained, is that AI token costs will have to be managed just like any other resource, offering an analogy to things like payroll or operating expenditure (OpEx), which is the day-to-day costs of running a business. “I think of it like…any other resource,” Mosseri said. “I have to decide how to deploy capacity to my different teams because I have a limited number of GPUs and CPUs and storage and RAM etc. I have to decide how to deploy OpEx for labeling budgets across my teams. I have to decide how to deploy payroll for headcount across my teams.” Token budgets will be the same, he added, noting that the cap per engineer would have to be proportional to the company’s trust in their ability to use the budget in an “ROI-positive” way. Meta doesn’t currently have token caps for any employee, Mosseri said, but he believes that their use could be healthy in the future. Further down the road, he expects token costs to come down as the AI model makers enter a pricing war to attract people to use their tools over their competitors. For now, the company has managed to rein in its token costs a bit by shutting down the “silly things” that it was doing, Mosseri noted — like that token spend leaderboard. “It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,” he said.
View
DeepMind CEO calls for an independent standards body to regulate frontier AI
In an X post on Tuesday morning, Google DeepMind CEO Demis Hassabis called for the creation of a new regulatory body to oversee frontier model releases. Titled“A Framework for Frontier AI and the Dawning of a New Age,”the post makes the case for a “standards body” modeled after theFinancial Industry Regulatory Authority(FINRA), which could test frontier models and develop best practices for their release. “Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release,” the post reads. “Once the assessment protocol is shown to be effective and robust, formalisation could quickly follow, meaning that Frontier Models would be required to pass it to be deployed in the US market. Labs would also work with the Standards Body to address any critical post-release vulnerabilities.”The proposed system would build on the ad hoc reviews performed by the U.S. government on Anthropic’s Mythos and OpenAI’s Sol. Those reviews drewsignificant criticismfor lack of technical expertise and opaque decision-making as to when a model could be released. Under Hassabis’ proposed regulator, those decisions could be handed off to a new organization, backed by the U.S. government but funded by the AI industry and operated independently. The prospect of AI regulation remains controversial for both the tech industry and the Trump administration. Most recently, White House AI advisor and a16z general partner Sriram Krishnandiscounted the possibilityof an AI regulator within the executive branch, saying “there will not be an FDA for AI.” Establishing the standards body as a self-regulatory organization like FINRA could be a way to address those concerns. Hassabis envisions the regulator being staffed by open source representatives and technical experts from within the industry, along with the financial backing from AI labs that would be necessary to retain them. They could even outsource some evaluations to the growing pool of AI safety groups that would be able to specialize in specific risks. “The strength of this approach is it would be technically focused, while at the same time supporting innovation and incentivising responsible behaviour,” Hassabis argues. “It is designed to keep up with the field’s acceleration and adapt to the biggest risks as they are identified, and could be ratcheted up if the seriousness of the situation demands.”
View
Google faces another AI training lawsuit from major publishers
A group of publishers and authors havefiled a class action lawsuitagainst Google, accusing the tech giant of using their copyrighted works to train its AI platform, Gemini. The group of plaintiffs, which includes Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E., also alleges that Google intentionally removed or changed copyright information on these works to “conceal… that its Gemini Models were trained on stolen materials,” according to the lawsuit. This lawsuit is just one of many complaints that publishers, authors, and other copyright holders have filed against AI companies such as Google, Meta, OpenAI, and Anthropic. While many of these lawsuits are still pending, two early court decisions in California havefavoredthe AI companies, ruling that the use of copyrighted works for AI training is considered “fair use” under U.S. copyright law that hasnot been updatedsince before the existence of the internet. Anthropic was, however, fined $1.5 billion for pirating the works it trained on, marking the largest payout in the history of U.S. copyright law. Around half a million writers were eligible for payments of at least $3,000. However, many authors opted out of receiving the settlement so that they could pursuefurther legal actionover AI training. The California judges’ decisions don’t bode well for how other courts may view the tech companies’ fair use defense, but the conflict is too nuanced for these rulings to establish an inarguable precedent. The lawsuit against Google was filed in the U.S. District Court for the Southern District of New York, giving a different judge the opportunity to weigh in. In the Google case, the publishers have a more nuanced, long-term relationship with the company. The lawsuit explains that publishers and authors have a long history of providing Google with copyrighted works for the specific purpose of making books searchable through Google Books. These search results do not allow users to view entire books. Instead, they provide access to short snippets of the book along with bibliographic information. The plaintiffs claim that Google trained Gemini on copies of these books, as well as books uploaded to the Google Play store, even though it never received permission to do so. “Google illegally copied works from all these scope-limited programs for AI training, knowing it lacked authorization to do so,” the lawsuit reads. The plaintiffs also cite an internal document from Google that allegedly states that using copyrighted books for AI training could be “highly problematic for Google” and might result in “$10Bs-$100Bs in potential fines.” Google did not immediately respond to a request for comment.
View
Meta Pulls Muse AI Image Feature Less Than a Week After User Backlash Highlights Privacy Risks
Meta introduced a new Muse Image feature last week, based on the firm's first AI model from the tech firm's Superintelligence Labs. The feature lets users reuse “publicly available” content, which has been posted by Instagram users. This AI-generated content can then be shared with others on various Meta-owned platforms, including WhatsApp and Facebook. This drew criticism from Instagram users, highlighting the privacy risks the newly added functionality poses. Now, the US-based company has updated the original blog post to announce that the feature has been removed. However, it did not address the highlighted privacy concerns.
View
Superhuman’s new auto-draft feature almost makes me like AI replies
Since the large language model (LLM) explosion started, companies have been trying to solve the problem of overflowing inboxes by using AI tocategorize emailsanddraft repliesthat sound like you. Email clientSuperhumanis launching a new version of its auto-draft feature that identifies important emails and creates draft replies that sound less robotic. Superhuman has attempted this in the past with features like instant replies and follow-up auto-drafts. However, a lot of those emails sounded like an overly enthusiastic AI salesperson, and I didn’t use them much. The new version of the auto-draft feature feels different. In the last few days, after gaining access to the beta, I have sent emails with little to no editing for some generated drafts. The app understands which emails might need replies and drafts a response based on your tone from previous conversations. It also generates two other variations that you might want to send instead. In my experience using the feature, I saw drafts agreeing to embargoes on a pitch to get more details, or confirming timing for a meeting that I could send with minimal edits. The feature also generated responses to emails asking for an authored post on TechCrunch, saying that I don’t handle that work. (TechCrunch does not accept authored posts.) The feature is far from perfect, though. By default, it often generated a positive response to a pitch, or agreed to a meeting at a post-midnight time. Thankfully, I could select another response from the other variations quickly and send it away. The feature learns from your usage and improves responses. For instance, after the midnight meeting debacle, when someone suggested a similar time, the feature generated a draft saying that the timing doesn’t work for me. I receive thousands of emails every month, partially thanks to AI making first drafts easier for others, like comms and PR professionals. I don’t have the confidence to hand over the reins to AI to handle my inbox completely, but this feature could help me respond to more people when I don’t need to type out long messages. Users can personalize emails by heading to Settings> Personalization and adding details about themselves and their role, along with adding files or links for more context. Superhuman’s co-founder, Rahul Vohra, said during the testing phase that 40% of auto-generated drafts were sent within one day, and 60% of those were sent without any manual editing. Vohra said that earlier features like Instant replies were built from older models like GPT-3.5, which were less intelligent or had a smaller context window. With this new implementation, the company is using an array of models. “Today, we are using a mixture of models to make this work. The actual writing is done by frontier models from both Anthropic and OpenAI. So we’re applying the maximum amount of intelligence and context to this that we possibly can to make the feature work,” Vohra said. Last year,Grammarly acquired Superhumanand then rebrandedthe company as Superhuman. Now, the company is building an assistant called Superhuman Go that spans platforms while carrying context over from one app to another.
View
Spotify expands its AI push with a ChatGPT-like music assistant
Spotify is taking another step to infuse AI technology into its listening experience, withTuesday’s newsthat Premium users will now be able to have interactive conversations with the app to choose what music or other audio they want to hear. The feature is initially available in the U.S., Ireland, and Sweden across iOS and Android devices for users 18 years old and above in English. It’s considered a beta release, meaning that things may not always work perfectly, Spotify says, but user feedback will help to improve the product. The company didn’t explicitly share more details about the AI technology under the hood in its announcement, but Spotify confirmed to TechCrunch that it uses a mix of its own AI technology and models from multiple providers, based on whatever is best for the task. The addition is the latest example of how Spotify has put AI technology to use to help people interact with the app’s extensive catalog of music, podcasts, and audiobooks. The company also offers tools like an AI DJ, which speaks in an AI voice that you can engage with directly, plus AI features for building playlists with prompts and those for connecting Spotify with third-party AI chatbots, like ChatGPT. Loading the player… The new feature extends the ability to chat with Spotify beyond the AI DJ experience, allowing users to talk to Spotify across the app’s Home and Now Playing views on mobile devices. Users can either type or speak to the app and have back-and-forth conversations to help them choose what to play next. Beyond that, Spotify says the app will also be able to chat with users about their listening history and can help them learn more about their favorite music or go deeper into podcasts or audiobooks. That means you could get into questions like what inspired a certain song, or dates of album releases, or even get suggestions of other artists you might like, based on what you’re playing. You can also ask about your own listening history, like when was the first time you played a certain track, or you could explore more into what sort of genres you’ve been streaming lately. In the announcement about the new feature, Spotify also offers a few suggestions as to how to use this interactive technology. For instance, you could ask Spotify to “play some artists I haven’t heard before,” then continue to shape that selection with follow-ups, like asking it to add a specific artist by name, or narrow the selection to just more recent tracks. You could also shape the request further by asking it to be “more upbeat,” or give it other directions. Plus, you can ask Spotify to save songs, add songs to your queue, or follow the artist via the new feature. The feature is rolling out now to the markets on mobile devices.
View
The real AI race may no longer be at the frontier
For several weeks this summer, the AI industry was fixated onAnthropic’s latest frontier modelsandWashington’s fight to controlwho was granted access to them. But while everyone was watching the frontier, developers kept building — and they weren’t waiting around for permission from the Anthropics and OpenAIs of the world. Chinese open-weight models accounted for41% of downloadson Hugging Face this spring, surpassing U.S. models. OnOpenRouter, the top six most popular models are all open models from Chinese firms including Tencent, Xiaomi, DeepSeek, MiniMax, and Z.ai. Anthropic’s Claude Opus 4.7 trails in seventh place, at the time of this writing. Anddata from Vercelshows thatopen weight modelsare absorbing much of the volume-heavy infrastructure of AI apps, while closed models operate as the higher-cost, premium layer. Open models handled nearly a third of AI requests on the platform in June. Those platforms only capture one slice of the AI ecosystem; in particular, they leave out sessions hosted by major labs, which likely account for the bulk of OpenAI and Anthropic’s usage. But open-source models’ large and growing share of the market raises a difficult question: How much do frontier models still matter if most production AI ends up running on cheaper, customizable alternatives? Some see the growth of open-source models as a sign that the most intelligent models may end up being used for only the most specialized use cases. “Maybe in a few years, the frontier models will be for experimenting and [for] some really high value tasks, and most of the production workloads will actually be powered either by private models within companies or by open source models,” Hugging Face CEO Clem Delangue said on a recentepisode of Equity. Loading the player… Hugging Face is a platform and developer community best known for hosting, sharing, and helping companies deploy open models. Delangue says Hugging Face’s customers and community members are increasingly touting the benefits of owning their own AI models rather than renting them, a trend that’s picked up steam in the cold light of day after getting the bill associated with thecost of scaling closed frontier models. “If you’re an AI company or a technology company, you don’t want to outsource your core capabilities to another company, to a black box API that you don’t control, don’t have any visibility on, and don’t really have any sort of ownership,” Delangue said. That shift, Delangue argues, is reflected in the activity happening on Hugging Face. A new repository is created every seven seconds on the platform, which hosts almost three million public models and one million public datasets, per Delangue. That points to a different picture than the “one model to rule them all,” he says. In reality, it looks more like companies using many different models, many of which are customized for their specific use case. Half of all Fortune 500 firms are using Hugging Face to deploy their own private models and open source models, he says. The growing popularity of open models coincides with a steady stream of increasingly capable releases from Chinese AI labs. Every few months, another Chinese AI company releases a powerful open-weight model that is cheaper to deploy and easier to customize than closed competitors, undercutting the economics of proprietary AI that U.S. firms have poured billions into. Most recently, Beijing-based AI company Z.ai released an open weight model called GLM-5.2 that excels at agentic coding and competes with Anthropic’s latest models on identifying security vulnerabilities. Delangue isn’t the only executive arguing that enterprises should avoid tying themselves to a single model provider. Microsoft CEO Satya Nadellarecently warnedagainst single provider lock-in, arguing that control of data should be a primary concern for enterprises using AI. “While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data,” Nadella said. “If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it’s imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop.” The rise of open models has also intensified a debate over whether increasingly capable models should be broadly available at all. Anthropic CEO Dario Amodei hasarguedthat scaling powerful open model weights could become dangerous because once they are released, they become difficult to control.Othershave argued that open models are easier to access by bad actors who could use them to spread disinformation or enact cyber or biological warfare. Delangue sees the tradeoff differently. “The biggest risk in AI is concentration of power,” Delangue said. “The way you make the world safer, in my opinion, is by leveling up the playing fields and creating transparency on these models.” Transparency means defenders can more easily “patch the cybersecurity risks that they already know open source models can exploit,” he said. The Hugging Face executive argues that keeping powerful models closed doesn’t eliminate the risks associated with advanced AI systems, in part because it’s easy to get past frontier model API guardrails and tosteal the weightsand disseminate them openly. Restricting powerful models, Delangue argues, simply concentrates the technology in the hands of a few companies while reducing transparency into how systems work. “You don’t really make it safe by keeping it behind closed doors for just a few players,” Delangue said. “You make it more dangerous because you create asymmetry of power and asymmetry of capabilities.”
View
Reflection inks $1B compute deal with Nebius
Reflection AI, a U.S. startup vying to developopen models, has signed a $1 billion compute deal with European AI infrastructure company, Nebius. Nebius, formerly the international arm of Russian tech giant Yandex, will provide Reflection access to Nvidia’s latest chips. The deal comes just a few weeks after the startup signed asimilar deal to access SpaceX’s computing resources, and mirrors several partnerships by AI firms as they race to secure compute for training and deploying their models. Along with its increasingly capable Chinese counterparts, Reflection is one of several open-weight AI model developers that have received ample attention lately as debate rages over the value of top-shelf, closed-source AI models — especially withdata retention concernssurging up, as well as government intervention. Just last month, the Trump administration pressuredAnthropicandOpenAIto restrict their most powerful new models, raising concerns that access toAI models could be taken away overnight. That, plus the release of more capable open models from China, has led to an increase in mainstream interest in open source AI. Reflection, currently valued at $8 billion, was founded in 2024 by two former Google DeepMind researchers. It has already raised close to $2.6 billion in funding from backers including Nvidia, Sequoia Capital, and Lightspeed Venture Partners. Shortly after securing a$2 billion investment from Nvidia, Nebius signed a five-year infrastructure dealwith Metaworth up to $27 billion. Last year, Nebius signed a multi-year deal with Microsoft worth up to $19.4 billion. TechCrunch has reached out to Reflection and Nebius for more information.
View
HCLTech's AI Infrastructure Play Can Redefine the Limits of Indian IT
HCLTech's AI strategy shows Indian IT is moving beyond software services, betting on compute infrastructure, sovereign AI, and specialised talent.
View
