AI News Archive: June 12, 2026 — Part 4
Sourced from 500+ daily AI sources, scored by relevance.
- General World Models
A research initiative for AI models
- Introducing Generalist Intelligence
Generalist Intelligence is a weekly intelligence briefing designed for tech’s most discerning professionals. It will arrive Friday mornings. Our hope is that it opens you up to perspectives you hadn’t considered and offers a look at fresh pockets of the future you might have missed.
- What comes after the AI spending binge: Caps, dashboards, and the search for ROI
Uber, Microsoft, and Meta are taming runaway AI budgets. The harder question is how to measure what AI spending actually produces
- Why Nokia is betting on AI-powered networks
Nokia helped define the mobile era. Now it's trying to shape what comes next. As artificial intelligence transforms how networks are built and managed, the Finnish telecom giant is betting on AI-driven infrastructure, 6G and a more software-centric future.
Score: 59🌐 MovesJun 12, 2026https://www.cnbc.com/video/2026/06/12/why-nokia-is-betting-on-ai-powered-networks.html - Google unveils DiffusionGemma, an AI model that breaks free of left-to-right processing
Google unveils DiffusionGemma, an AI model that breaks free of left-to-right processing InfoWorld
- Siri AI is coming to newer Apple devices only, here’s the full list
Siri AI is a major upgrade and big highlight of iOS 27, macOS Golden Gate, watchOS 27, and more. But only newer Apple devices will support the overhauled Siri. Here’s the full list of Siri AI-compatible hardware across iPhone, iPad, Mac, Apple Watch, and Apple Vision Pro.
Score: 58🌐 MovesJun 12, 2026https://9to5mac.com/2026/06/12/siri-ai-is-coming-to-newer-apple-devices-only-heres-the-full-list/ - Tiny ‘crawling’ robot solves Japan’s moon lander mystery
Tiny ‘crawling’ robot solves Japan’s moon lander mystery
- ASUS’s New Lineup Doubles Down On AI
ASUS’s New Lineup Doubles Down On AI india.entrepreneur.com
Score: 58🌐 MovesJun 12, 2026https://india.entrepreneur.com/business-news/asuss-new-lineup-doubles-down-on-ai - AI Is Coming for Jobs. The Question Is Whether Governments Are Paying Attention.
Entrepreneur Marco Riedesser has spent his career building automation systems, which is precisely why his views on AI and jobs stand out. In a conversation that moves from coding and robotics to universal income and purpose, he argues that the real challenge may not be AI itself, but whether society is prepared for how work could change.
- AI Is Changing Cybersecurity Faster than Most Businesses Realize
Artificial Intelligence is transforming cybersecurity at an unprecedented pace, creating new opportunities for businesses while simultaneously equipping cybercriminals with increasingly powerful tools. As organizations accelerate AI adoption, cybersecurity leaders are warning that the threat landscape is evolving faster than many companies realize. In a conversation with Mukul Kumar, Managing Partner, Claracon AI, he highlighted that […] The post AI Is Changing Cybersecurity Faster than Most Businesses Realize appeared first on CXOToday.com .
- How Much Does It (Really) Cost to Use Claude Fable, GPT-5.5, and Gemini 3.5 Flash?
As if anyone is supposed to know what a million tokens adds up to.
Score: 58🌐 MovesJun 12, 2026https://gizmodo.com/how-much-does-it-really-cost-to-use-claude-fable-gpt-5-5-and-gemini-3-5-flash-2000770837 - Moonshot AI Launches Kimi Work, a Local Desktop Agent Reportedly Running on Kimi K2.6 With a 300-Sub-Agent Agent Swarm
Moonshot AI Launches Kimi Work, a Local Desktop Agent Reportedly Running on Kimi K2.6 With a 300-Sub-Agent Agent Swarm MarkTechPost
- AI Cloud Service Provider E2E Networks Debuts On BSE Mainboard
NSE-listed E2E Networks’ stock made its debut on BSE’s Mainboard exchange today post securing the necessary approvals from the exchanges.…
Score: 58🌐 MovesJun 12, 2026https://inc42.com/buzz/ai-cloud-service-provider-e2e-networks-debuts-on-bse-mainboard/ - Frontier AI models could be an adversary's force multiplier
Frontier AI models such as Anthropic Claude, Mythos , and OpenAI Daybreak fundamentally alter the cybersecurity equation by compressing the time, skill, and scale required to discover and exploit vulnerabilities. A single adversary can now automate reconnaissance, generate exploit variants, analyse source code, weaponise misconfigurations, and adapt phishing or social engineering campaigns at machine speed. For CISOs, the problem is no longer just “AI adoption risk” but the rise of AI-amplified adversaries capable of iterating faster than traditional defense cycles. Combating frontier AI model risks and threats In this evolving landscape, organisations must address the risks and threats posed by frontier AI models by combining human expertise with AI-assisted defenses and treating security as a continuously adaptive function rather than a periodic exercise. CISOs need to establish new policies, operational procedures, and governance models not only to defend against the misuse of frontier AI but also to strategically leverage these technologies to strengthen the organization’s overall security posture. Let us explore how CISOs can adapt to manage and mitigate the emerging risks associated with frontier AI models. The key elements of a strategy to securely combat risks arising from frontier AI models such as Anthropic Claude Mythos. Continuous exposure management CISOs need to shift from traditional monitoring to continuous exposure management. In the age of AI, quarterly assessments are too slow when AI can continuously analyse attack surfaces. Security teams should prioritise continuous asset discovery, external attack surface management, automated configuration validation, and rapid patch orchestration tailored to AI entities. Equally important is reducing the “blast radius” of inevitable compromise through zero-trust segmentation, least-privilege access, short-lived credentials, and robust identity governance. The assumption should be: if AI can find it, it will eventually be exploited . AI-aware defence engineering This mechanism reflects the integration of AI-focused threat modeling into the SDLC and infrastructure design. Development pipelines should include AI-assisted code review, secret scanning, dependency risk analysis, and automated policy validation before deployment. Focus on securing high-risk AI infrastructure components, such as APIs, plugins, agents, MCP-style integrations, and AI-connected workflows, which significantly expand the attack surface. Defenders need behavioral analytics to detect abnormal automation patterns, autonomous reconnaissance behavior, and machine-speed lateral movement. Read more about Claude Mythos Anthropic's Claude Mythos has generated buzz and alarm among CIOs and CISOs, who fear the model could expose vulnerabilities and drive unprecedented levels of hacking . As AI tools such as Claude Mythos Preview can speed vulnerability discovery for attackers, CIOs are automating detection and response to keep pace . Claude Mythos has the potential to enhance global cyber security or undermine it by becoming a weapon in the hands of threat actors . AI surface governance and reducing breach risk CISOs must recognise that AI surface governance and resilience are critical strategic requirements, not compliance exercises. Security policies must govern the use of frontier models, Shadow AI adoption, prompt usage analysis, third-party AI integrations, and agent permissions. CISOs must adopt a shift-left strategy for vulnerability discovery, using the same class of AI-powered tools, i.e., frontier AI models, to uncover the attack surface adversaries could exploit. At the same time, organisations should prepare operationally for AI-enabled breaches: tabletop exercises, AI-red-team simulations, supply-chain compromise scenarios, and incident response plans that assume adversaries can adapt dynamically during an intrusion. The key mindset shift is that frontier AI models are accelerants that reshape the speed, scale, and asymmetry of cyber conflict. Rapid response assuming AI speed disclosure The window between vulnerability discovery and exploitation is narrowing. CISOs must understand patch and response process needs and assume that a critical vulnerability may be weaponised within 24 hours of disclosure, or even sooner. Relying on slow patch cycles, manual triage, or periodic security reviews is not viable when adversaries can automate reconnaissance, weaponisation, and exploitation at machine speed. The time demands rapid-response security models that include pre-positioned response playbooks, AI-assisted prioritisation, and resilient architectures capable of quickly containing compromise. In practice, CISOs must assume that once a weakness becomes visible, AI-enabled adversaries can rapidly operationalise it before traditional defences can react. Reshaping privileged access for AI entities We are witnessing the evolution of AI solutions that use active agents to interact with APIs, infrastructure, workflows, and enterprise data. CISOs must reshape the privilege-access model for dynamic AI entities, such as agents. Organszations require tightly scoped, identity-aware, and time-bound access models tailored to the AI entities accessing frontier AI models. This means applying zero-trust principles to AI agents, continuously validating their actions, monitoring behavioral deviations, and enforcing granular controls over which data, systems, and operations they can access. With the advent of frontier AI models and AI agents, privileged access management is no longer just about securing human administrators; it is about governing machine-driven entities operating at scale and speed. The need of the hour: CISO mindset shift CISOs' practical line of thought: stop planning for the attacker you knew and start planning for the attacker that frontier models enable . That attacker is faster, more contextually aware, more persistent, and more scalable than anything the security industry has faced. CISOs who adapt most quickly to manage the AI attack surface will lead enterprise security in the frontier-model era. Those who treat this as an incremental update to existing frameworks will find that the gap between their defenses and the threat has quietly become insurmountable. CISOs need to internalise this speed asymmetry before building any response strategy. Read more in this series John Bruce, Quorum Cyber: Claude Mythos forces the conversation on defensive AI. Martin Riley, Bridewell: Mythos is turning up the heat on risk, not rewriting the rules.
Score: 57🌐 MovesJun 12, 2026https://www.computerweekly.com/opinion/Frontier-AI-models-could-be-an-adversarys-force-multiplier - Xiaomi’s latest AI coding tool claims to outperform Claude Code on complex tasks
Xiaomi’s latest AI coding tool claims to outperform Claude Code on complex tasks
- Daily 5 report for June 12: Artificial intelligence and robots will redefine car design and the future of factory work
Daily 5 report for June 12: Artificial intelligence and robots will redefine car design and the future of factory work Automotive News
Score: 57🌐 MovesJun 12, 2026https://www.autonews.com/newsletters/daily-5/an-daily-5-ai-auto-manufacturing-job-losses-0612/ - The Truth About AI Data Centers And Why It Matters
Not every discussion about data centers is grounded in fact, and many people have limited visibility into how these facilities operate or what AI workloads require.
Score: 57🌐 MovesJun 12, 2026https://www.forbes.com/councils/forbestechcouncil/2026/06/12/the-truth-about-ai-data-centers-and-why-it-matters/ - Kimi K2.7 Code now available on AI Gateway
Kimi K2.7 Code from Moonshot AI is now available on AI Gateway . K2.7 Code is a coding model built for long-horizon programming tasks, generalizing across scenarios including frontend development, DevOps, and performance optimization. The model has a native multimodal architecture that supports text and vision input, and always runs in thinking mode. To use K2.7 Code, set model to moonshotai/kimi-k2.7-code in the AI SDK : Pass an image alongside a prompt to use the model's multimodal input: AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. It includes built-in custom reporting , Zero Data Retention support , budgets for API keys , and more. AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests. Learn more about AI Gateway , view the AI Gateway model leaderboard or try it in our model playground . Read more
- AI will create more jobs than it displaces, says Microsoft India chief
In a blog post, Kumar said rapid technological change is shortening the shelf life of skills, making continuous learning essential.
- UW researchers built AI agents that quickly estimate electronic devices’ carbon footprints
UW researchers built AI agents that quickly estimate electronic devices’ carbon footprints EurekAlert!
- Why most enterprise AI programs fail — and how to turn them around
Enterprises have invested billions in AI, yet many programs remain stuck in proof-of-concept, with models that rarely influence decisions. The challenge isn’t technology — it’s operating models, fragmented data, governance gaps and organizational misalignment. To succeed, AI must be treated as a strategic capability that drives measurable business value to gain competitive advantage, not just a technical tool. This article explores the key structural and organizational factors that enable AI to move from pilots to enterprise impact The 5 structural barriers to enterprise AI success Most enterprise AI programs stall due to five structural barriers that go beyond technology. Understanding these challenges is the first step toward turning AI from a pilot experiment into a true enterprise capability. 1. Fragmented data ecosystems AI cannot scale when data is siloed across functions. Models may succeed in pilots, but disconnected systems prevent enterprise-wide deployment. Breaking down silos with unified data platforms enables consistent, reusable pipelines and lays the foundation for scalable AI. In fact, architectural research regarding why your AI is failing and how a smarter data architecture can fix it shows that up to 90% of enterprise data goes unused for analytics, leaving models contextually devoid of real operational meaning. 2. Lack of clear business ownership Many AI initiatives originate within technology teams rather than business units. This often shifts the focus to capability creation — building models, platforms or experiments — rather than solving real business problems. Without strong business ownership, AI remains a technology exercise. Leaders must anchor AI programs to measurable business outcomes. 3. Pilot-driven culture and limited institutional readiness Generative AI has fuelled experimentation, but running dozens of pilots is not transformation. Small-scale experiments rarely account for enterprise-scale performance, and the biggest mistake is scaling models instead of decision systems. This discrepancy is underscored by recent data showing that while nearly every enterprise is investing in AI, only 5% say their data is ready to support it at production scale. True AI adoption requires the capabilities, culture and trust to embed AI into decision-making. 4. Human-in-the-loop (HITL) architecture designs HITL systems are often introduced as risk controls during early pilots. But when embedded into production workflows, they can limit scalability. Enterprises should use HITL selectively — for exception handling — rather than as a permanent dependency that slows adoption. 5. Governance and risk concerns Without proper guardrails, enterprises hesitate to operationalize AI at scale. Pilots can be controlled, but scaling AI requires governance structures, risk management and new ways of working. Organizations that fail to build these frameworks often see stalled programs and missed opportunities. The difference between AI experiments and AI transformation Many organizations mistake running machine learning models for AI transformation. True transformation occurs when AI is embedded into enterprise decision-making and operational processes, not just used for pilots. AI models may perform well in experiments or limited deployments, but they rarely scale without end-to-end integration across business workflows, data systems and governance frameworks. Successful AI transformation requires a shift from: Experimentation → Enterprise platforms Technology focus → Business outcomes Isolated models → Integrated decision systems The goal is not simply to build pilots or deploy models. It is to embed intelligence into core business processes, enabling decisions that are faster, more accurate and more consistent across the enterprise. The leadership model needed to scale AI AI transformation is not purely a technical challenge — it requires a leadership and operating model that demands alignment across governance, business strategy and operational execution. When programs begin to drift, leadership must understand how to rescue failing AI initiatives by gaining clear visibility into orchestration gaps rather than falling into sunk-cost thinking. Key elements include: Enterprise AI governance. Establish clear policies for model lifecycle management, risk and compliance. Governance is critical not just for regulatory compliance, but for creating the trust and accountability needed to integrate AI into core business processes. Cross-functional collaboration. AI initiatives must bring together data teams, technology leaders, business stakeholders, process excellence teams and organizational design experts. Success depends on designing AI end-to-end—from data pipelines to decision workflows — rather than simply deploying models. AI product ownership. Treat AI capabilities as products with defined owners, measurable outcomes and roadmaps for continuous improvement. Enterprises should hire AI product managers who bridge business objectives, data science capabilities and operational deployment, ensuring that AI delivers tangible value at scale. Building an enterprise AI platform Scaling AI across the enterprise requires platform thinking. The most successful organizations treat AI infrastructure as a shared capability, similar to cloud platforms, enabling teams across business units to innovate faster and more efficiently. Rather than building isolated models or point solutions, enterprises should develop: Reusable data pipelines that provide consistent, high-quality data across applications Shared model infrastructure to accelerate experimentation and deployment Standardized deployment frameworks that simplify scaling and integration Enterprise AI governance to ensure compliance, risk management and operational accountability This approach reduces duplication, accelerates time-to-value and ensures that AI becomes a strategic enabler rather than a collection of technology experiments. From technology initiative to enterprise capability AI succeeds only when it becomes a core business capability — not a standalone technology project. Enterprises that treat AI strategically achieve measurable impact and reshape decision-making. How to turn AI pilots into enterprise capability: Anchor AI to business Value. Identify high-impact problems, set KPIs tied to revenue, efficiency or customer experience, and ensure business units own the initiative. Build a shared enterprise AI platform. Develop reusable data pipelines, shared model infrastructure and standardized deployment frameworks. Treat AI as a shared capability, enabling cross-team collaboration, faster experimentation and scalable deployment. Create cross-functional AI teams. Form integrated squads that include data scientists, technology leaders, business stakeholders, process experts, organizational design and operational teams. True AI transformation requires collaboration across functions to design end-to-end solutions, not just isolated models. Establish governance and product ownership. Define enterprise AI governance for risk, compliance and operational oversight. Appoint AI product managers to align business strategy, data science and operations. Create roadmaps and measurable outcomes to ensure continuous value delivery. When executed correctly, AI moves beyond experimentation. It embeds intelligence into core business processes, drives faster and more consistent decisions, and delivers enterprise-wide impact. Key takeaway The challenge is not building AI models — it’s building organizations that know how to use them. Enterprises that win with AI don’t focus on the number of models; they redesign how decisions are made, embed intelligence into core processes and deliver measurable business outcomes. By focusing on value, shared platforms, cross-functional collaboration and governance, organizations can move AI from stalled experiments to enterprise-wide decision-making. Organizations that succeed will be those that rethink their operating models, align leadership around outcomes and build scalable AI platforms that power enterprise decision-making. This article is published as part of the Foundry Expert Contributor Network. Want to join?
Score: 56🌐 MovesJun 12, 2026https://www.cio.com/article/4184158/why-most-enterprise-ai-programs-fail-and-how-to-turn-them-around.html - How BBC Eye built a multi-agent AI system to sift through ten thousand Russian social media posts
How BBC Eye built a multi-agent AI system to sift through ten thousand Russian social media posts reutersinstitute.politics.ox.ac.uk
- Mint Explainer: Why AI data centres are the new goldmine for Indian law firms
As global tech giants spearhead a massive AI infrastructure expansion, law firms are securing a long-term pipeline of advisory work spanning land, tech, and power.
- AI Needs Ajinomoto’s Film, but Its CEO Won’t Raise Prices Just Because
The world’s hunger for artificial-intelligence chips is straining the industry’s ability to produce the materials they rely on, creating new pressure points that could slow the boom.
- Waymo: Aim to Become “World’s Most Trusted Driver” and “New Reference Model for Human Collision Avoidance”
With the launch of the 2026 FIFA World Cup, Waymo is on a publishing spree this week. Yesterday, it published a scientifically oriented article on a “new reference model for human collision avoidance.” This is the result of joint research with TU Delft, and a resulting paper was published in ... [continued] The post Waymo: Aim to Become “World’s Most Trusted Driver” and “New Reference Model for Human Collision Avoidance” appeared first on CleanTechnica .
- Israeli startup unveils laser system it says can kill drones in seconds
Israeli startup unveils laser system it says can kill drones in seconds Breaking Defense
Score: 55🌐 MovesJun 12, 2026https://breakingdefense.com/2026/06/israeli-startup-unveils-laser-system-it-says-can-kill-drones-in-seconds/ - AI tracks missing hydrogen atoms in crystals with 97% success rate
Artificial intelligence is often used to generate images. In research, specialized AI models are used for scientific applications—for example, to predict the positions of atoms in materials. The MatterGen model developed by Microsoft can generate complex crystal structures from just a few pieces of information—which atoms should be present and in what proportions—and researchers can then use these structures for computer simulations of new materials.
- Apple explains the goal of iOS 27’s new AI features in Photos
The Photos app is a central focus of iOS 27, adding new features like Reframe and Extend. In a new interview with Tyler Stalman , two Apple Photos executives went in-depth on the new features and what they mean for photographers.
Score: 55🌐 MovesJun 12, 2026https://9to5mac.com/2026/06/12/apple-explains-the-goal-of-ios-27s-new-ai-features-in-photos/ - Why ChatGPT’s new ad strategy is the best thing that’s ever happened to Apple Intelligence
Apple Intelligence may still be coming of age, but it has its monetization plan sorted.
Score: 55🌐 MovesJun 12, 2026https://www.androidauthority.com/chatgpt-ads-good-for-apple-intelligence-why-3676436/ - AI chatbots are not your lawyer. Protect attorney‑client confidentiality.
AI chatbots are not your lawyer. Protect attorney‑client confidentiality. AJC.com
Score: 55🌐 MovesJun 12, 2026https://www.ajc.com/opinion/2026/06/ai-chatbots-are-not-your-lawyer-protect-attorneyclient-confidentiality/ - PNC's solution to rising AI costs: build its own AI
The bank has decided it doesn't want to be at the mercy of tech companies and the price they demand for services, and is building its own "AI factory."
Score: 55🌐 MovesJun 12, 2026https://www.americanbanker.com/news/pncs-solution-to-rising-ai-costs-build-its-own-ai - Spain bets on AI and automation to power its fruit and vegetable industry
Spain is Switzerland’s top supplier of fruit and vegetables and the country is now turning to new technology to transform the sector. Swissinfo travelled to El Ejido in the province of Almería to see how farmers and companies are investing millions in artificial intelligence, machinery and robotics to boost precision, cut labour costs and stay competitive in the European market.
- New malware campaign tricks AI scanners with fake nuclear weapon prompts — malicious code triggers safety failsafes so scanners skip the payload
Hades malware campaign now tricks AI bots into not scanning development packages, as prompts for bio- and nuclear weapons trigger failsafe mechanisms.
- Middle East boards lead world on AI governance and future readiness, global study finds
Middle East boards lead world on AI governance and future readiness, global study finds Arabian Business
- AI is a runaway train. Businesses risk repeating a decade of failed digital transformation
In the wrong hands, AI adoption can be awful, in particular because so many of the people trying to roll it out fail to consider the workers who are being asked to use it, writes Karima-Catherine Goundiam
- Mozilla's CEO Knows You Might Not Want AI in Firefox
We talked to the browser maker's new boss about privacy, choice and his vision for the internet's future.
Score: 55🌐 MovesJun 12, 2026https://www.cnet.com/tech/services-and-software/mozillas-ceo-knows-you-might-not-want-ai-in-firefox/ - The AI adoption spending spree is over. Time to focus on value.
IT leaders and CFOs are starting to push back on unrestrained AI spending within their organizations, with many enterprises now looking for ways to get better value out of their automation tools, observers say. In recent months, several companies have blown through AI token budgets while encouraging employees to experiment with the technology. Several companies have deployed AI token usage leaderboards to advance adoption, but the practice has led to employee tokenmaxxing , with one Disney employee interacting with the Claude AI 460,000 times in a nine-day period. Over at Uber, employees burned through the company’s entire 2026 AI budget in four months, prompting executives to cap AI coding assistant use. At the same time, Anthropic and other major AI vendors are rolling out new metered pricing models . Now, to counter runaway token use, some companies have begun to focus on “ valuemaxxing ,” by combining visibility, governance, and financial accountability to create better returns on their AI spending. Value as a new metric The focus at many enterprises in past years has been using AI to increase productivity, but value is increasingly becoming an important metric, says Becky Trevino , chief product officer at FinOps vendor Flexera. “Before, the value was, ‘Let’s just deploy it to as many people as possible,’” she says. “It’s now quickly becoming, ‘Let’s be smart about this because we have a set amount of tokens.’” AI budget surprises have become a serious talking point among customers of the technology, she adds. “The metric for success is not if I have 99% of my organization using Microsoft Copilot,” Trevino says. “It’s now, show me the money. Show me what’s changing in your organization.” FinOps companies such as Flexera aren’t the only organizations pushing AI customers to focus on value. In early June, the Linux Foundation announced the new Tokenomics Foundation to establish open industry standards, benchmarks, and best practices for the economics of AI infrastructure. The bill comes due A reckoning about AI value and cost is overdue, some experts say. IT leaders who aren’t worried about the price of AI services are underestimating the coming problems, says Cliff Jurkiewicz , vice president of global strategy at HR-focused AI vendor Phenom “AI is underpriced right now, and that’s masking the real cost,” he adds. “Behind every prompt is expensive infrastructure — data centers and energy — and providers aren’t absorbing that forever.” There seems to be a casual approach to AI token use at many organizations, he says, with the current cheap access phase encouraging overconsumption. “There’s very little discipline around what actually drives business value versus what’s just convenience,” Jurkiewicz says. “Every interaction has a cost, input and output, and these systems tend to overdeliver, which means you’re paying for more than you need. Multiply that across an entire workforce, and costs scale fast.” As Anthropic and OpenAI both appear to be headed toward IPOs, pricing models will change, he predicts. Smart AI users will respond by limiting use, he adds. “As pricing shifts to metered models, companies are going to get a wake-up call,” Jurkiewicz says. “The winners will be the ones that put guardrails in place early by prioritizing high-impact use cases and treating AI like a finite, strategic resource, not an unlimited utility.” IT leaders need to recognize that costs can vary widely between AI models, adds Andy Sen , CTO of IT services marketplace AppDirect. Many companies are allowing employees to use models without considering the cost, he says. “A lot of companies are turning on AI features everywhere, riding the productivity wave until they are blindsided by a huge bill,” he adds. Encourage savings Sen encourages IT leaders to research the costs of individual models and to encourage employees to use the most economic choice available. IT leaders should consider turning off premium features for small tasks, such as writing emails. “The key is understanding the difference in cost between some of these models, because there are some that can be one hundred times more expensive than others,” he says. “Instead of getting an answer in fifteen seconds, you’re getting it in five seconds, and for most work, that’s not really worth it.” CIOs can also help employees use AI more economically by pointing out the workers who are using it well, Sen says. “If you want to see more people leveraging AI across the board, point to the people who are already using it successfully,” he says. “Role models can be a great way to show people how they can be more effective at their job and encourage AI use.” Flexera’s Trevino recommends that organizations consider FinOps services and prioritize which departments have more access to AI tools. “If you have a limited amount of budget and you have 20 departments, which of those departments matter the most?” she says. “Allocate most of your AI budget there. If you know three of those 20 are the ones that are you need to drive growth or competitiveness for your business, then you should really ensure that those three business units have the access they need.”
Score: 55🌐 MovesJun 12, 2026https://www.cio.com/article/4183263/the-ai-adoption-spree-is-over-time-to-focus-on-value.html - Google Releases Gemini-SQL2: Gemini 3.1 Pro Text-to-SQL Scores 80.04% on BIRD Single-Model Leaderboard
Google Releases Gemini-SQL2: Gemini 3.1 Pro Text-to-SQL Scores 80.04% on BIRD Single-Model Leaderboard MarkTechPost
- What enterprises are now demanding from AI, and who is actually meeting the mark
The AI market has moved past potential. Enterprises are now evaluating products on a single metric: measurable business impact. The ET Most Innovative AI Product Awards 2026 recognise the solutions that are proving it.
- CoRover AI builds AI without the internet, and here’s why it works better
CoRover AI builds AI without the internet, and here’s why it works better YourStory.com
Score: 54🌐 MovesJun 12, 2026https://yourstory.com/2026/06/corover-ai-offline-edge-ai-deployment-ankush-sabharwal-devsparks-2026 - Publicis Sapient launches Sapient Sustain for autonomous IT operations
As enterprises accelerate AI adoption, their IT environments are becoming increasingly complex. Hybrid and multi-cloud architectures, distributed applications, legacy systems, and growing volumes of operational data have created challenges that […] The post Publicis Sapient launches Sapient Sustain for autonomous IT operations appeared first on Express Computer .
- Corning Is Riding High on the AI Boom—and Planning Ahead in Case It Goes Bust
CEO Wendell Weeks remembers the dot-com crash and other hard times, and those lessons have taught him to hedge even the most optimistic data-center bets.
Score: 54🌐 MovesJun 12, 2026https://www.wsj.com/tech/ai/corning-wendell-weeks-ai-deals-data-centers-82ad26d7?mod=rss_Technology - 5 foundations for reshaping the future of education and AI
The post 5 foundations for reshaping the future of education and AI appeared first on Source .
- Building Box AI: How an Enterprise Content Platform Went AI-Native with Deep Agents
Box AI's journey to becoming AI-native with Deep Agents
- Siri AI is all Apple; it just needed Google to get there
Apple’s executives have been taking questions, hosting seminars, seemingly working around the clock to stress one very important thing: Apple is not using a white label version of Google Gemini to make Siri AI happen. They just pooled resources to get there. The new Siri AI is faster, more accurate, offers powerful contextual capabilities and shows how Apple has leap-frogged into a good peer position in an AI race critics felt it had already lost. Its market scale — even without the EU — is huge. For most consumers, Apple Intelligence and Siri will continue to be their primary/first engagement with artificial intelligence on a device. Getting there took a lot of work, and Apple needed Google to get it done. Though there is still some confusion about what that means, Apple’s software chief tried to explain it this week. “We use none of the models that Google deploys to their customers, nor do we use the infrastructure and means by which they employ models to their customers,” Craig Federighi said in a presentation at WWDC . Apple is not even using Google Search as the foundation of its system, he said. “This is the amount of Google Assistant we use,” Federighi said, pointing at an empty chart. “Nothing.” Apple not Gemini, Siri AI is not Google’s What makes this hard to understand is that we all know Apple partnered with Google to build Siri AI; back in January, we were told the next generation of Apple Foundation Models would be based on Gemini models and cloud technology. So, how can we have moved from partnership hero to usage zero? The answer is, we didn’t. What happened is that Apple built its new Apple Frontier Models (AFMs) (the AI inside Siri AI) by training them using proprietary Apple data and reinforcement learning and then refined those models using “outputs from Google’s Gemini Frontier models.” In other words, Apple used Google Gemini to help improve its own models, which means the models themselves, the AI in Siri AI, are Apple’s — but they were trained with help from Gemini. They are not white label iterations of those Google models. Apple also hit a second snag. Its very best model (AFM 3 Cloud Pro) requires more processor power to run than Apple could deliver using its own cloud-hosted Private Cloud Compute servers . Now, we know Apple doesn’t like using other people’s stuff. But it’s a realistic company that understands it sometimes must, and just as it uses AWS to support some of its services, it moved to adopt Google cloud services and Nvidia processors to drive the most demanding requests. Apple A question of trust Apple also developed a technologi c al solution that means it can claim the interaction remains just as private as if it were run on your device. Apple has made it possible for independent security experts to confirm this and says it is the only company that can deploy software on those servers, with strong security to ensure your device only interacts with those servers when you want it to. So far, no one has broken this protection. I came across an interesting report in which Tekonyx Founder and Chief Research Officer Sid Nag explained the significance of Apple finding a way to expand its Private cloud Compute infrastructure beyond its own data centers. “Apple is effectively arguing that trust in AI systems should come from cryptographic and architectural guarantees rather than trust in the cloud provider itself,” he told Fierce . Where enterprises have faced a choice between access to powerful AI or privacy, Apple has introduced a new solution he called “portable trust.” This could conceivably become a new IaaS offering from the companies involved over time. Working together for the benefit of all So, while Apple’s models were built with Google’s help, and while its most advanced models run with support from Google and Nvidia, the models are Apple’s alone. It’s good for Google, of course. Apple is paying for this use, which helps the search giant claw back some of the value of its massive, muti-billion-dollar AI infrastructure investment . It’s not clear how much Apple is paying; earlier this year, the $1 billion figure was bandied around. But Apple’s decision to tie usage to iCloud subscriptions in some hitherto undisclosed way hints that the deal may also see some token-based usage charges on top of the basic Apple fee. I’ve not come across any details, but that’s what I surmise based on the size of Apple’s ecosystem and the growing realization of how quickly users can consume AI capacity . What AI models is Apple running? AFM 3 Cloud Pro is one of five Apple Frontier Models driving Siri AI. Here’s how Apple describes those five models : On-device models AFM 3 Core , the next generation of Apple’s 3-billion-parameter dense model. You’ll use this for basic text generation, summaries, conversational replies, all the standard uses. It can also handle indexing, search, App Intents, basic dictation and contextual awareness. AFM 3 Core Advanced , the most powerful on-device model. This is what makes Siri’s voice match mood or context, does all the high-accuracy dictation, and can process various data to handle tasks across different apps. It’s the AI driving your more involved Siri conversations. AFM Core Advanced is impressive in its own right, because Apple has managed to cram a 20-billion-parameter model onto a smartphone. It has done this by using a sparse architecture, which means it activates just 1 to 4 billion parameters at a time depending on the request. It is, however, only available to Apple’s most powerful systems — iPhone 17 Pro, iPhone 17 Pro Max, iPhone Air, M4 or later iPad, M3 or later Mac with 12GB+ memory. Server-based models AFM 3 Cloud , which Apple calls its server-side workhorse, optimized for speed, efficiency, and performance. ADM 3 Cloud (Image ), for image generation and editing, which unlocks advanced photo-editing tools, the all-new Image Playground, and more. AFM 3 Cloud Pro , the most capable server-based model, which powers the most demanding use cases, like agentic tool use and complex reasoning. All five were built using the same common foundation, which was then specialized to reflect the proposed use of that model. It’s interesting to look at the human evaluation tests Apple ran to test how well these models performed ; they demonstrate impressive improvement on the company’s original models, effectively justifying the decision to work with Gemini. You can follow me on social media! Join me on BlueSky , LinkedIn , Mastodon and subscribe to The Core .
Score: 53🌐 MovesJun 12, 2026https://www.computerworld.com/article/4184484/siri-ai-is-all-apple-it-just-needed-google-to-get-there.html - Two Turing Award Winners Confront the Theoretical Black Hole Behind AGI at BAAI Conference
Whitfield Diffie and Andrew Barto, two Turing Award laureates, delivered keynote speeches at the 8th BAAI Conference in Beijing, converging on a shared concern: the fundamental theoretical challenges behind ensuring AGI safety and alignment.
- The Judgment Tax: How AI Agents Are Rewriting UI Process Automation
Agents can handle work requiring judgment and unstructured information, not just the clean rules-based tasks RPA was designed for.
- TensorWave COO Piotr Tomasik sees 'dynamic change' in AI investment
TensorWave COO Piotr Tomasik sees 'dynamic change' in AI investment
Score: 52🌐 MovesJun 12, 2026https://qz.com/tensorwave-coo-piotr-tomasik-ai-investment-dynamic-change - Kimi K2.7-Code cuts thinking tokens 30% — but practitioners say the benchmarks don't check out
Moonshot AI released Kimi K2.7-Code this week, an open-source update to its K2 coding model family, claiming leaner reasoning and double-digit performance gains. K2.7-Code is built on the same trillion-parameter mixture-of-experts architecture as its p redecessor K2.6 , and drops in via an OpenAI-compatible API — which matters for teams already running K2.6 in production gateways. When K2.6 launched in April, it topped OpenRouter's weekly LLM leaderboard — a ranking based on actual API routing decisions by developers, not self-reported benchmark scores. Moonshot AI says K2.7-Code addresses what it calls "overthinking," reducing thinking-token usage by 30% compared to K2.6 — a number that would directly affect inference costs for teams running agentic workflows. Whether that efficiency gain holds on independent benchmarks is a question practitioners have already started raising publicly. What Kimi K2.7-Code is K2.7-Code is released under a Modified MIT license, with weights available on HuggingFace. The model is deployable via vLLM or SGLang. It runs exclusively in thinking mode and does not support temperature adjustment — Moonshot AI has fixed it at 1.0, meaning teams cannot tune output determinism the way they might with other models. The core change from K2.6 is how the model generates low-level code. Where K2.6 produced implementations by wrapping existing libraries and routing through established frameworks, K2.7-Code authors implementations directly. Moonshot AI says this produces more reliable generalization across Rust, Go and Python, and across task types including frontend development, DevOps and performance optimization. On benchmark performance, Moonshot AI claims gains of 21.8% on Kimi Code Bench v2, 11% on Program Bench and 31.5% on MLS Bench Lite. All three are proprietary benchmarks run by Moonshot AI. The model has not been submitted to DeepSWE, an independent coding benchmark that produces a 70-point spread across models — compared to SWE-Bench Pro's 30-point spread — making it a more discriminating signal for teams configuring model routing systems. More honest, weaker for it The picture from outside Moonshot's own benchmarks is more complicated. Researcher Elliot Arledge ran K2.7-Code against K2.6 and Claude Fable 5 on KernelBench-Hard, a public benchmark focused on GPU kernel optimization, and published his full run logs at kernelbench.com. "K2.7 is more honest but not more capable," Arledge wrote on X . On five of six problems, K2.7-Code produced real authored Triton kernels where K2.6 had used library wrappers. Two of those kernels failed on the model's own bugs. The MoE kernel result regressed from K2.6's score of 0.222 to 0.157. "Fable, for reference, tops every cell it doesn't honestly fail," Arledge wrote. Sugumaran Balasubramaniyan, a developer who built a model-task-router for the Hermes Agent platform using DeepSWE as his reference signal, responded publicly to the K2.7-Code release and challenged Moonshot AI directly on the benchmark choices. "Respectfully, every model 'improves' double digits on its own test suite," Balasubramaniyan wrote on X . He noted that K2.6 scored 24% on DeepSWE, tied with GPT-5.4-mini, and asked whether Moonshot AI would submit K2.7-Code to the same benchmark. Balasubramaniyan said it took 13 review rounds to get the benchmark data right for his router and that he would route coding tasks to K2.7-Code if the independent numbers hold up. What this means for enterprises The token efficiency gain is immediately usable. Teams running K2.6 in production can swap in K2.7-Code via the OpenAI-compatible API and expect lower inference costs on agentic workflows without an architecture change. The 30% thinking-token reduction is Moonshot's own number, but the integration path is low-risk enough to test against your own workloads before committing. The practical question is whether those efficiency gains hold on a team's own task distribution. Running K2.7-Code against your own workloads before adjusting gateway weights is the low-risk path to finding out.
- The AI industry's platform trap is starting to look a lot like Microsoft's
Anthropic is throttling its new Mythos model for certain tasks while building apps that directly compete with its largest customers. Customers, partners, and investors are pushing back. The article The AI industry's platform trap is starting to look a lot like Microsoft's appeared first on The Decoder .
Score: 52🌐 MovesJun 12, 2026https://the-decoder.com/the-ai-industrys-platform-trap-is-starting-to-look-a-lot-like-microsofts/