AI News Archive: July 3, 2026 — Part 7
Sourced from 500+ daily AI sources, scored by relevance.
- Alibaba Orders Employees to Remove All Anthropic Products, Ban Effective July 10
Alibaba internally bans Anthropic's Claude models and agent products, effective July 10, following accusations of model distillation attacks.
- Alibaba bans Claude Code after Anthropic is caught tracking Chinese users with hidden code
Alibaba has banned its employees from using Claude Code, Anthropic’s AI-powered coding agent, after security researchers discovered that the tool contained hidden code designed to identify Chinese users. The ban, effective 10 July, follows weeks of escalating conflict between the two companies over allegations that Alibaba stole Anthropic’s AI capabilities through industrial-scale distillation. “As Claude Code was […] This story continues at The Next Web
- OpenAI is offering the Trump administration a 5% stake to seed an AI wealth fund
OpenAI CEO Sam Altman reportedly discussed the potential government stake with Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent
- OpenAI might give the US government a 5% stake to fund citizen payouts
OpenAI proposes a 5% stake in the US government to help fund citizen payouts.
- OpenAI proposes handing U.S. government a 5% stake, report says
OpenAI proposes handing U.S. government a 5% stake, report says The Japan Times
- OpenAI Woos Trump Administration as Investor
OpenAI Woos Trump Administration as Investor Time Magazine
- Nvidia is launching a revenue-sharing model to give AI startups access to its hardware
The chipmaker will earn a share of cloud revenue from partners, with two data center operators named as its first participants
- Man charged with manslaughter over Tesla crash originally blamed on car’s self-driving mode
Tesla said Michael Butler disabled his car’s self-driving mode before it plowed into Martha Avila’s home in June A man whose Tesla Model 3 was allegedly in self-driving mode when it crashed into a home near Houston and killed a 76-year-old woman inside recently has been jailed on a count of manslaughter. Michael Butler’s arrest in the 19 June death of Martha Avila was announced late on Wednesday in a Facebook post by the sheriff of Harris county, Texas, Ed Gonzalez. Continue reading...
- Tesla driver charged with manslaughter after crash into Texas home
Tesla driver charged with manslaughter after crash into Texas home USA Today
- UK fintech Starling to axe 130 roles in AI-powered simplification drive
Starling is gearing up to cut some 130 roles in a bid to streamline its operations following a fall in profit over the last year. The London-based fintech told staff this week it would restructure its banking and tech operations as it leveraged AI across operations ahead of an influx of new projects. It said [...]
- Elon Musk caps Tesla staff’s AI spending
Elon Musk caps Tesla staff’s AI spending The Telegraph
- Chipmaker Kioxia ships next-gen memory samples as AI boom fuels dramatic comeback
The growth of AI has fueled a remarkable turnaround for Kioxia, which was previously seen as an example of Japan's chipmaking struggles but whose stock price has surged more than sevenfold this year to a market capitalisation over $250 billion, exceeding Toyota Motor's.
- HCLTech Bags Landmark $1.14 Billion Enterprise AI Contract
HCLTech Bags Landmark $1.14 Billion Enterprise AI Contract YourStory.com
- Mark Zuckerberg admits Meta’s AI ambitions face major delays — Here's why
Mark Zuckerberg admits Meta’s AI ambitions face major delays — Here's why YourStory.com
- AI agent tech progressing slower than expected, says Mark Zuckerberg
Zuckerberg added that a company reorganization that included major job cuts was not as "clean" as it could have been and that executives had miscalculated on the timing of the changes
- Meta’s Zuckerberg says AI agent tech progressing slower than expected
Meta’s Zuckerberg says AI agent tech progressing slower than expected
- Meta's AI agent push is moving slower than Zuckerberg planned
Mark Zuckerberg admitted to weaknesses in the company's restructuring during an internal town hall. The AI agents Meta reorganized around are progressing slower than planned, Zuckerberg said. His AI chief, meanwhile, painted a rosier picture. The article Meta's AI agent push is moving slower than Zuckerberg planned appeared first on The Decoder .
- Mark Zuckerberg admits the AI restructuring isnt going all that great, report says
Meta CEO Mark Zuckerberg said AI agents aren't progressing as quickly as he had expected, new report says.
- Mark Zuckerberg says Meta’s agentic AI efforts aren’t progressing as fast as he had hoped
Meta Platforms Inc. Chief Executive Mark Zuckerberg told employees at an internal town hall meeting that the company’s work on artificial intelligence agents hasn’t progressed as quickly as he had hoped. It’s a rare admission from the founder that not everything is going to plan as it tries to turn its multibillion-dollar investments into AI […] The post Mark Zuckerberg says Meta’s agentic AI efforts aren’t progressing as fast as he had hoped appeared first on SiliconANGLE .
- What Meta Said About Slow Progress on AI Agents
What Meta Said About Slow Progress on AI Agents Barron's
- Meta AI Fears Ease Despite Zuckerberg’s Disappointment in Agents
Meta AI Fears Ease Despite Zuckerberg’s Disappointment in Agents Barron's
- Anthropic is in talks with Samsung to develop an AI chip
Anthropic is in talks with Samsung to develop an AI chip YourStory.com
- Osloq
An AI agent that reproduces GitHub issues for you
- Anthropic in talks with Samsung for a potential AI chip collaboration: Report
Anthropic in talks with Samsung for a potential AI chip collaboration: Report
- Samsung Electronics, Anthropic in AI chip talks: Report
Samsung Electronics, Anthropic in AI chip talks: Report 매일경제
- Anthropic reportedly in talks with Samsung to manufacture custom AI chip
Anthropic PBC is reportedly in talks with Samsung Electronics Co. to produce a custom artificial intelligence chip. The Information today cited sources as saying that the discussions are in an early stage. According to the publication, Anthropic has not yet finalized key details such as what workloads the processor will run and its performance. Some […] The post Anthropic reportedly in talks with Samsung to manufacture custom AI chip appeared first on SiliconANGLE .
- Time has come to look at separate AI legislation, says IT Secy S Krishnan
India appears set to move towards a dedicated regulatory framework for artificial intelligence, with IT Secretary S Krishnan on Friday saying the time has come to look at a separate AI regulation. Krishnan noted that while existing legal provisions have so far been adequate in addressing initial concerns on issues like deepfakes and AI-generated synthetic content, an "additional regulation or law may be needed". "It is a conversation which has commenced, and my Minister (IT Minister Ashwini Vaishnaw) and I have both been on record earlier that we will look at AI regulation when the time is right, and it appears that the time is getting right, and we will start looking at it," Krishnan said. He added: "We have used the IT rules, and other provisions of existing law to address various concerns that AI raises, but now, probably the time has come to look at a separate legislation." Asked about the timelines for bringing out a new AI regulation, the IT secretary said: "As Ministry, at a
- Govt may look at a separate legal framework for AI: Secretary
India is considering a dedicated legal framework for Artificial Intelligence, a significant departure from its previous stance of using existing laws. The Ministry of Electronics and IT believes the time is right to draft specific AI legislation. This move signals a potential shift in the government's approach to regulating the rapidly evolving technology, aiming to balance innovation with necessary oversight.
- India Signals AI Policy Shift: Time is right for dedicated law, says MeitY
MeitY Secretary says while existing legal provisions have so far been adequate, an “additional regulation or law may be needed” now
- Forget coding: Meta's new Pocket app lets AI build games from your ideas
Forget coding: Meta's new Pocket app lets AI build games from your ideas Gulf News
- Microsoft Copilot Cowork and the Rise of the AI-Native Work
How the shift from chat to execution is redefining what it means to work effectively with AI. For the past few years, being “good at AI” mostly meant being good at prompting. The professional advantage came from knowing how to ask better questions, structure better instructions, and iterate until a chatbot produced a useful answer. That model is already starting to feel dated. The next major shift in AI is not just from weaker models to smarter models. It is from chat to execution . And that shift changes the job description of the person at the keyboard. With Microsoft Copilot Cowork , AI is no longer just an advisor that responds with text. It becomes a work partner that can understand an outcome, create a plan, reason across tools and files, and help carry work forward across Microsoft 365 workflows. Instead of asking Copilot for a single draft or summary, users can begin delegating broader, multi-step work. That is a very different relationship with AI. From Prompting to Delegation Traditional AI chat works like a helpful consultant. You describe a task, receive guidance, and then do the actual work yourself. You copy content into documents, search folders, update slides, send emails, manage calendars, and coordinate follow-ups. Execution-focused AI changes that pattern. With Copilot Cowork, the interaction moves closer to delegation. You describe the outcome: prepare for a customer meeting, build a weekly update, organize research, summarize action items, create a draft deck, or coordinate follow-ups. Copilot Cowork can then break the work into steps, operate across Microsoft 365 apps and organizational context, and keep you in the loop as progress happens. The skill shifts from “Can I write the perfect prompt?” to “Can I define the right outcome, provide the right context, set the right boundaries, and review the result?” That is the foundation of the AI-native professional. Who Copilot Cowork is for Cowork is not just for developers. It is aimed at anyone whose work is spread across files, apps, documents, inboxes, calendars, and repeated routines. It tends to help most for: Knowledge workers producing reports, research, summaries, and spreadsheets Founders and solopreneurs who need leverage without adding headcount Marketers running content workflows, scheduling, and social listening Students and educators managing notes, deadlines, and research material Ops and admin teams handling files, inboxes, scheduling, and coordination Non-developers who want automation without writing a line of code The qualifier is simple. If you only want an LLM to tell you what to do, chat is enough. If you want an LLM to do the work alongside you, Cowork is the better fit. What makes Copilot Cowork different Copilot Cowork executes complex, long-running, multi-tool tasks. You define the work and Cowork runs it end-to-end and returns a completed result, not just a draft or a recommendation. Cowork is designed to be more accurate, more secure, and lower cost than other offerings. Five things make that true: Cloud hosting means files are not stored locally, security is strongly enforced, and your tasks keep running even when your laptop is off. Native Work IQ support grounds every task in the systems your business already runs on, so the work reflects real context. Enterprise-grade security and compliance ensure Cowork operates within your Microsoft 365 trust boundary, with protections that align to your organization’s existing policies and controls. Multi-model design lets you run the models a task needs, so capability scales with the work as more models become available. Lower cost from a runtime that efficiently finds the right information and tools, model choice that matches the right model to each task, and billing that charges you only for what you use. When comparing the cost per prompt between Copilot Cowork and Claude Cowork with their Microsoft 365 connector, testing showed that Copilot Cowork on average was 30–40% cheaper. Our newest model, Cowork 1, will be a secure, fine-tuned model releasing in the coming weeks, post-trained to handle tasks at a substantially lower cost. You are not locked into one model with Copilot — you can use the most efficient model or frontier models. Pricing model Copilot Cowork requires the Microsoft 365 Copilot User Subscription License (USL). Users are then billed for Cowork on a usage-based basis, with charges determined by the tasks they run. The Microsoft 365 Copilot USL includes a complete AI productivity experience: Copilot Chat; Copilot in Word, Excel, PowerPoint, Outlook, and Teams, the Work IQ context engine; a multi-model system offering frontier intelligence; pre-built agents like Researcher and Analyst; and custom agents built with Agent Builder. All of that for a predictable per user per month fee. Copilot Cowork adds a whole new way of working: an agentic system designed for complex, long-running, multi-tool tasks. You are billed on a usage basis, denominated in Copilot Credits, and the price for each task is calculated from four inputs: model use, context retrieval, tool calls, and runtime. In my customer meetings over the last few months, the most frequent question I’ve heard is how to budget for Cowork given its variable pricing model. From usage during the Frontier program, we observed three common task patterns: light, medium, and heavy tasks. Light tasks use a small number of knowledge sources, apply limited reasoning, and produce one or fewer outputs. Medium tasks draw on multiple sources, apply structured reasoning, and generate two or more outputs. Heavy tasks aggregate broadly, apply deep reasoning, and produce many outputs. We also identified four user personas with distinct usage patterns across these task types. When you combine users by persona with their mix of light, medium, and heavy tasks — and apply a price per prompt — you get a flexible cost model that can help you create a cost estimate you can refine over time. At its core, the model multiplies the number of users in each segment by their expected prompt volume across light, medium, and heavy tasks, applies the cost per prompt type, and sums the total. If you want to model this yourself, you can download a simple spreadsheet here . These estimates assume Anthropic Opus 4.8. At general availability, Copilot Cowork runs on Anthropic models, including Opus 4.8 and Sonnet 4.6. In Frontier, customers can use GPT 5.5, with Cowork 1 coming soon. Cowork 1 will deliver optimal cost, quality, and enterprise-grade use, including removing model bias. It’s designed to handle everyday Copilot tasks at a substantially lower cost, making it a strong option for cost-sensitive workloads. For more detail on the pricing model, refer to this Microsoft Learn article . Cost management Usage-based billing makes it critical to track the value and ROI you get from AI. Long-running, agentic work can create a lot of value for your business but also requires significant compute resources. Three factors will continue to reduce cost over time: models will get cheaper, Cowork will get better at matching models to tasks, and both context retrieval and tool use will become more efficient. In the meantime, cost management is one of the most important new capability areas in the general availability of Cowork, and we’re shipping features across three themes: control, visibility, and efficiency. Control: Customers decide when Cowork turns on, who gets access, and how much can be spent. Cowork is off by default. Admins decide when to enable Cowork in their tenant and who gets access. Spending limits at the tenant, group, and user levels. Admins create scoped billing policies and define budgets, including user-level caps set inside group policies. Customizable usage alerts. Admins set the thresholds that matter for their organization and groups, and choose who gets notified when spend crosses them. User-initiated credit requests. When a user needs additional credits to complete a task, they can request them from inside Cowork. Visibility: Customers see what’s being used, and what each task costs. Usage reporting at the tenant, group, and user levels. Admins see usage broken down by user, group, and feature, with clear accountability across the organization. User-level pricing for each task, in credits. Users see what each task costs as they run it. Efficiency: Customers have options to manage cost. Two payment options: PayGo and P3. Customers choose pay-as-you-go for flexibility, or P3 to commit to a usage volume in advance in exchange for a discount. PayGo is priced at $0.01 per Copilot Credit. Model choice in Frontier. Where multiple models are available, customers can use the model picker to manage cost-per-task. Billing for Copilot Cowork begins today. Before usage ramps, admins can set spending limits and allocate budgets using the cost management controls described above. Tenants that had at least one user in the Frontier program (March 30 to June 16) who used Cowork during that period get a grace period on Cowork usage and will not be billed until July 1, 2026, to support the transition. For additional details on our cost management features, please see this Microsoft Learn article . AI-Native Workflows Start Small The biggest mistake professionals make with AI is trying to automate everything at once. A better approach is to start with small, bounded workflows. Pick something repetitive, low-risk, and easy to verify. For example: Prepare a weekly team update from recent notes and emails Summarize action items from meetings and draft follow-ups Organize research into a structured briefing document Turn spreadsheet insights into a short executive summary Create a first-pass PowerPoint from a Word document Collect customer context before a meeting Draft recurring status reports using the same structure each week The goal is not to replace your work in one step. The goal is to identify repeatable work and gradually hand off bounded pieces of execution. Over time, those small workflows become reusable systems. How to build your AI-native workflow and to get started with Copilot Cowork Before you begin, make sure you have: Microsoft 365 Copilot access: An active Microsoft 365 Copilot license assigned to your account. A modern browser: Microsoft Edge or Google Chrome recommended. Cowork available: Cowork is enabled in your Microsoft 365 Copilot environment. Usage-based billing: Usage-based and Cowork billing has been enabled. Anthropic enabled in tenant: Cowork uses Anthropic models as a subprocessor to ensure secure and responsible use of Anthropic models within your organization. Details about the integration can be found at Anthropic as a subprocessor for Microsoft Online Services . Cowork works in your browser at m365.cloud.microsoft , in Outlook and Teams, in the Microsoft 365 Copilot desktop app for Windows and Mac, and in the Microsoft 365 Copilot mobile app for iOS and Android. Open Cowork Open Microsoft 365 Copilot . Select Cowork in the top toggle next to Chat. When the Cowork homepage loads, you’ll have access to: A chat input where you can describe what you need, along with any recent tasks you can pick up where you left off. Search: instantly search and revisit your previous tasks, so your past work is always at your fingertips. Scheduled: view, edit, reschedule, and clean up your scheduled tasks without hunting through menus. Customize: add your own personal skills to Cowork, and discover and add plugins to extend what Cowork can do. Model picker: choose the model that works best for your task, or let Cowork decide. Start your first conversation Type what you want done in the chat input, or select one of the suggested prompts such as Catch me up, Organize my inbox, or Prep for a meeting. You can enter up to 250,000 characters, so feel free to include plenty of detail. Attach files if needed. Select the Add attachments button (the + sign) to choose from Add work context (files, people, meetings), Upload images and files from your device, or Attach cloud files from OneDrive, SharePoint, or Teams. You can also drag and drop files onto the input area. (Optional) Use voice input by selecting the microphone icon to speak your request instead of typing. Send your message by selecting Send. Watch Cowork work After you send your message, Cowork begins processing your request. Here’s what appears in the chat as each step happens: A thinking indicator: Appears while Cowork figures out the best approach. Skill messages: Show when Cowork loads a skill it needs, such as “Preparing to compose emails” or “Preparing to create Word documents.” Tool steps: Show exactly what Cowork is doing at each stage, such as Composing your email or Creating your presentation . The response streams in: In real time, so you can follow along as Cowork builds your result. Approve actions Before Cowork takes an action on your behalf, like sending an email, posting a Teams message, or scheduling a meeting, it asks for your permission. The approval dialog shows a preview of what Cowork plans to do and a button labeled with the specific action. Approve button (for example, Send, Post, or Create): Allow this specific action to proceed. Always allow: Select the dropdown arrow next to the action button to allow the action and skip future approval prompts for similar actions in the current conversation. Cancel: Stop this action. Sometimes Cowork asks you a question to clarify your request. When this happens, you see a set of choices you can select from, or you can type your own answer. Select Skip if you’d rather not answer. Review your results When Cowork finishes, any files it created appear in the side panel on the right. From there you can: Download individual files to your device, or select Download All to download every file as a single zip archive. Preview files directly in the browser. Supported formats include PDF, Microsoft 365 documents (Word, Excel, PowerPoint), Markdown, code files, images, CSV, HTML, and email. Open files in OneDrive and directly in the online version of PowerPoint. The side panel also shows a progress bar with the percentage of tasks complete and the skills Cowork used. You can select thumbs down on responses that missed the mark, or thumbs up on ones that were helpful. View your created tasks The task view lets you select any task to open it and resume the conversation. You can filter tasks by those that need input, those that are still in progress, completed tasks, and scheduled tasks. The New Skill: Workflow Design AI-native professionals will not simply be better prompt writers. They will be better workflow designers. They will know how to define outcomes clearly, provide relevant context, set guardrails, inspect plans, verify outputs, and convert repeated tasks into reusable processes. They will understand when to use chat, when to use Copilot inside an app, when to use an agent, and when to use Copilot Cowork for broader multi-step work. That distinction matters. A quick question belongs in chat. A document edit belongs inside Word. A spreadsheet analysis belongs inside Excel. A repeated business process may belong with an agent. But an end-to-end workflow that spans apps, files, meetings, and deliverables is where Copilot Cowork becomes more interesting. The Real Takeaway Becoming an “AI-native professional” is less about being a prompt engineering wizard and more about being a great delegator. The skill set has shifted. It is no longer about “Can I write the perfect prompt?” It is about: Can I define a clear outcome? Can I bound the workspace safely? Can I inspect the AI’s plan? Can I turn a one-off task into a repeatable workflow? We aren’t talking about replacing your judgment; we’re talking about freeing you from the manual operator role. The professionals who thrive in this era won’t be the ones who let AI do everything blindly. They will be the ones who design better workflows, keep a steady hand on the oversight, and let AI carry the weight of execution. Stop asking your AI for advice. Start asking it to get to work. Microsoft Copilot Cowork and the Rise of the AI-Native Work was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.
- SAP cuts hiring and travel to fund AI
SAP is limiting hiring and travel spending to help pay for its AI transformation. The tech giant will “exclusively focus new hiring on selected profiles only, mainly core Al roles, that are critical for our long-term success,” staff were reportedly told in an internal email. The email also said that internal travel, unless it is related to AI development, will be suspended, and that the company is looking at ways to cut other spending with suppliers, Bloomberg reported . An SAP spokesperson confirmed the report, telling CIO , “SAP continually reviews its investments to ensure resources are focused on the areas that will drive long-term customer value and innovation. As part of this approach, we are prioritizing investments in AI-related capabilities, talent, and technologies while applying greater discipline to hiring, external spending, and internal travel. Customer-facing activities and critical AI initiatives remain fully supported.” This is yet another part of the SAP’s efforts to accelerate its focus on AI, including its digital assistant, Joule. Earlier this week, CEO Christian Klein took on direct responsibility for most of its AI development teams. In March, Klein had passed oversight of sales, delivery, service, and support to the new Customer Value Group under executive board member Thomas Saueressig, now chief customer officer. Customers need to see value Terra Higginson , principal research director at Info-Tech Research Group, said that while SAP needs AI adoption to support its strategy and justify its investments, “customers still need to see a clearer value proposition before they commit more budget or operational attention.” Like many software companies, SAP is facing multiple pressures, she said: SaaS valuations remain significantly below prior-cycle highs, AI is expensive to build, operate, and scale, and the commercial payoff from AI remains uncertain. “This is not a time for lavish spending,” she said. “SAP needs to be disciplined about where it invests, focusing on areas that create clear competitive differentiation. Joule has been underwhelming so far, yet I hear that SAP is pushing users hard to turn it on. That creates a real tension.” Reshaping the workforce AI is also behind SAP’s attempt to avoid layoffs like those that occurred during its 2024 restructuring . The company is encouraging employees to invent “more valuable jobs,” assisted by the new technologies, The New York Times reported . The report said that in the not too distant future, CEO Klein is expecting to see not a smaller workforce, but a very different one; he is not sure whether there will be any people coding software in two or three years. Jason Andersen , VP and principal analyst at Moor Insights & Strategy, said that when workers use AI a lot, it changes how they interact with their daily tasks. For example, software engineers are now doing more security and testing tasks as their time is freed up from coding. “But, this rebalancing of work is the big challenge that hasn’t been worked out yet,” he said. “That is the missing link in this whole future-of-work story and how that translates to today’s worker. And that has three major mitigating factors that at least temporarily disrupt good intentions.” First, he said, AI tends to be a personal productivity enhancer and is not yet team friendly. Second, “There’s an argument that AI will enable us to be so much more productive since we can now do those things that we could not before. Except that those things never really had a strong enough case to begin with. So, will we use this new capacity to do that or just cut the budget?” And thirdly, he pointed out, “macro views get sorted out in years and decades, not months and quarters. AI will change jobs, and, if you believe the research suggesting automation actually increases the number of jobs over time (which I do), it will work itself out.” SAP and other companies must do these things to stay competitive over the long run, he said, but, “In the short run, a lot of companies will have to scale back to move forward, which is of little consolation to those impacted.”
- SAP tightens hiring, travel as investments in AI zoom
German software giant SAP is implementing hiring freezes, travel restrictions, and tighter control over external spending to fuel its significant investments in artificial intelligence. The company is prioritizing core AI roles and strategic acquisitions to bolster its AI capabilities. While internal travel is paused, essential trips for AI initiatives and customer engagement will continue, reflecting SAP's commitment to its AI-driven future.
- Meta’s AI chief says new Muse Spark update will sharpen coding, agentic AI
Meta is set to launch a new Muse Spark model with stronger coding and agentic capabilities, as Chief AI Officer Alexandr Wang touted the update as a step toward closing the gap with rival AI platforms and expanding the company’s enterprise AI ambitions. “..Our next Muse Spark update is coming soon. Big improvements in coding and agentic capabilities to be more competitive with other leading models,” Wang wrote in a post on X in an attempt to clarify CEO Mark Zuckerberg’s comments about the slow progress in AI agent development made during a company townhall. In the same townhall, Wang said that the next Muse Spark update, codenamed Watermelon, which uses far more compute than its predecessor, has already caught up with OpenAI’s flagship GPT 5.5 model , according to a Business Insider report that cited anonymous sources. What the update means for enterprises The stronger coding and agentic capabilities in Watermelon, according to Pareekh Jain , principal analyst at Pareekh Consulting, could benefit enterprises. “A strong Meta model would increase competition, lower AI costs, and give enterprises another alternative to OpenAI and Anthropic,” Jain said. “If offered as an open-weight or low-cost model, it could make AI coding assistants more affordable while improving data control and reducing vendor lock-in,” Jain added. The analyst was referring to a broader shift in enterprise software development, where wider adoption of AI coding assistants has coincided with mounting cost and availability pressures, as GPU shortages, high model licensing fees, and inference costs make access to the most capable coding models increasingly expensive. The timing of the Muse Spark update and Meta’s recent acquisitions, including its efforts to acquire Manus , has fueled speculation that Meta might introduce its own AI-assisted application development platform or vibe coding tool. “It seems, especially with these updates, Meta wants to move beyond foundation models and become a platform for building AI-native applications and agents,” said Charlie Dai , principal analyst at Forrester. “While the status of Manus remains uncertain due to reported regulatory challenges, initiatives such as Pocket , although consumer-facing, indicate Meta’s interest in lowering the barriers to creating AI-native software. The more important opportunity, however, is enterprise adoption: enabling business users to build workflow automations, agents, and lightweight applications with less technical expertise,” Dai added. The analyst’s comments also align with Meta’s broader push into the enterprise AI market. Meta is reportedly developing plans for new cloud infrastructure business lines that would sell access to AI computing power and models. Enterprise opportunity comes with execution hurdles However, enterprise adoption might not come easy for Meta, analysts cautioned. “Meta must prove superior real-world coding quality, reliable agent execution, strong security and governance, and a vibrant developer ecosystem,” Dai said. “In addition, outside North America, geopolitical and regulatory considerations are increasingly shaping model choices and creating opportunities for alternatives. Meta needs compelling customer outcomes, strong local partnerships, and sustained innovation that resonates with developers and enterprises,” Dai added. The new model, according to Wang, will be rolled out soon via Meta AI and a new API. The article originally appeared on InfoWorld .
- Meta reuses old RAM in new servers with custom bridge chip
With the cost of new RAM soaring, Meta has found a thrifty way to reuse older memory in newer servers. The performance of about 40% of Meta’s millions of servers is limited by a lack of memory, the company said — but it has a surplus of older DIMMs from decommissioned servers, because RAM chips can last about twice as long as the rest of the machine. To profit from this imbalance, it developed a custom Computer Express Link (CXL) chip it calls Vistara, and associated software, to decouple older memory from server memory channels, enabling its reuse in new machines alongside their native memory. Using the older RAM with the CXL interface doesn’t significantly affect performance — although it would have done if the older DIMMs were plugged straight into newer servers. Kudos to tech site The Register for noticing the development, which Meta described in a technical paper: Vistara: Making CXL Real — Full Path from ASIC Design and OS Support to Hyperscale Deployment,” setting out how the new technology works . There is a particular need to be thrifty right now, given the current state of the market. Last year, users were warned that memory prices could double by the end of 2026, while the RAM shortage could last until 2027 . This week, Apple suggested using cheap Chinese chips , a move that may well be frowned on by the Trump administration. The Meta development may prove to be an efficient way forward.
- AI race weakens climate pledges at Google and Amazon
Google and Amazon this week reported sharp increases in greenhouse gas emissions, driven by the frantic construction of artificial intelligence infrastructure that is pushing the tech giants further from their carbon neutrality pledges. Amazon's emissions, published on Wednesday, have risen 58 percent over the same period, and more than 16 percent last year, despite a pledge to reach carbon neutrality by 2040.
- AI race weakens climate pledges at Google, Amazon
Google and Amazon this week reported sharp increases in greenhouse gas emissions, driven by the frantic construction of artificial intelligence infrastructure that is pushing the tech giants further from their carbon-neutrality pledges.
- Will the next Messi be a robot?
While the world's eyes are on Lionel Messi at the FIFA World Cup, another soccer tournament unfolds in Incheon — with robot players. RoboCup 2026 kicked off Thursday in Incheon's Songdo Convensia, showcasing a wide range of artificial intelligence robots capable of performing complicated tasks such as soccer and household chores. Often referred to as the "World Cup of Robotics," RoboCup has been held annually since its first event in Nagoya, Japan, in 1997. Thousands of competitors bring robots
- Move over, Messi! Robot footballers thrill crowds
Move over, Messi! Robot footballers thrill crowds Gulf News
- Move over, Messi! Robot footballers thrill crowds in South Korea
Thirty seconds before kick-off, humanoid robot footballers in red and blue jerseys await the referee's signal in the South Korean port city of Incheon.
- Humanoid robots showcase football skills
Humanoid robots compete in football matches at RoboCup in Incheon, South Korea, showcasing advances in robotics as spectators watch the action.
- AI Romance Scam Impersonating Dubai Prince Ensnares Victims
AI Romance Scam Impersonating Dubai Prince Ensnares Victims Barron's
- Free AI courses for 60,000 accounting, finance professionals in S’pore under national programme
Free AI courses for 60,000 accounting, finance professionals in S’pore under national programme The Straits Times
- I've Seen the Scary Headlines About AI. Here's Why I'm Still a Techno-Optimist
I've Seen the Scary Headlines About AI. Here's Why I'm Still a Techno-Optimist PCMag UK
- I've Seen the Scary Headlines About AI. Here's Why I'm Still a Techno-Optimist
I've Seen the Scary Headlines About AI. Here's Why I'm Still a Techno-Optimist PCMag Australia
- World shares rally after Dow hits a record, as some AI shares bounce back
World shares rally after Dow hits a record, as some AI shares bounce back Dallas News
- AI Agents Are Creating a New Enterprise Security Gap
Five independent security disclosures in a single week point to the same gap: AI agent permissions, not AI agent capabilities, are the problem enterprises haven't solved. The post AI Agents Are Creating a New Enterprise Security Gap appeared first on TechRepublic .
- Yet another research breaks the hype bubble for AI browsers serving serious security flaws
Researchers tested seven popular AI browsers and found four vulnerable to attacks that trick the AI agent into handing over personal data.
- Some agentic AI browsers may come with major cybersecurity risks
In the last year or so, artificial intelligence companies have rolled out a spate of web browsers equipped with AI agents. A user might ask one of these agents to plan a vacation, and it will open browser tabs to research routes and restaurants, then make reservations and add events to the user's calendar. How well it does any of this varies.
- Robots can now 'see' touch thanks to a new color-changing tactile sensor
Engineers at Queen Mary University of London have built a new color-changing tactile sensor, which allows robots to "see" and touch in real-time. The novel idea was invented by Giacomo Sasso, a postdoctoral researcher at the School of Engineering and Materials Science at Queen Mary University of London, and it works by transforming invisible forces into dynamic color patterns. This enables high-resolution maps of contact, strain and pressure to emerge instantly.