Latest AI News
Gujarat Taps IBM, IAIRO to establish Industrial AI Centre of Excellence
Envisioned as a ‘living lab’, the Industrial AI CoE will support the development, testing, and adoption of industrial AI applications
View
CoreWeave Launches ARIA to Help Researchers Find Hidden Patterns in AI Experiments
ARIA analyses thousands of experiment runs in minutes, surfaces hidden patterns and recommends improvements to accelerate AI innovation
View
Delhi to Roll Out AI-Enabled PUCC 3.0 Ahead of Winter Pollution Season: Report
The new PUCC 3.0 system will use AI, geotagging and encrypted data transmission to curb fraudulent vehicle emission certificates.
View
MongoDB to Upskill 2 Mn Builders in India by 2030, Unveils AI Retrieval Tools
The company announced Voyage Context 4, Hybrid Search, and Native Reranking, saying the technologies work together to improve retrieval quality.
View
Vibe coding platform Base44 launches own model as AI startups seek defensibility
Base44, the vibe coding platform thatWix acquired for $80 millionjust one year ago — when the company was barely six moths old and had a team of eight — has started rolling out its own AI model to support its users in creating apps with natural language. The move comes as the discussion in AI circles has intensified over whether frontier models are best suited for all use cases. A related question is whether businesses built on top of someone else’s models are truly defensible long-term. The latest move of Base44, based in the Bay Area, speaks to both. While its custom LLM is only just rolling out, Base44 hopes that it will eventually outperform frontier models. According to its founder, Maor Shlomo, “training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency.” At first glance, this could be a way to stay ahead of competitors such asSwedish startup Lovable, which reachedunicorn statusin its Series A round last summer and thatrelies on external LLMs. However, Shlomo expects that others will train their own models — “at least the players that have gotten enough scale and velocity to have enough data.” According to Jonathan Userovici, a general partner at VC firm Headline — whose portfolio includes AI companies like Mistral AI, but not Base44 — data is one of three key ingredients of defensibility for AI startups, alongside distribution and tech stack. The upshot is that players with strong brands are now leaning into their data and infrastructure to increase their defensibility, and Base44 fits that pattern. The company says the first iteration of its LLM, Base1, was developed and trained on a dataset generated from “tens of millions of real user interactions on the platform.” This dataset will keep on growing with the company; but so will its rivals’. The bigger competition may not be vibe-coding startups at all but instead come from frontier AI labs that are getting closer to Base44’s home turf — Cursor and Grok’s parent company xAI now bothbelong to SpaceX, and Claude Code has become a vibe coding player in its own right. This gives Anthropic and other foundational AI providers access to data and feedback loops they can use to improve models for app creation, but Shlomo thinks specialization gives Base44 a leg up. “Models are progressing, but they’ll stay very general in what they can do,” he predicted. Userovici, for his part, cautioned against underestimating frontier models, citing the example of the legal tech startup Harvey, which abandoned plans to train its own model. He doesn’t expect applied AI companies to become frontier labs en masse but frames Base44’s move in a broader context — one in which inference costs have become a meaningful part of the equation. That cost pressure, Userovici says, has driven change that enterprise customers are now demanding. “They don’t necessarily see a [return on investment] when using the latest models for all use cases, so an entire infrastructure is being set up to do orchestration and optimization to select the right models for them so that costs don’t skyrocket while maintaining the same or similar performance across the majority of use cases.” Enterprise companies still are a minority among the audience of the vibe coding platforms, but they represent a growing share of platform revenue, and users of all sizes are starting to express concerns over the cost of using AI. Base44’s decision to develop its own LLM stemmed from multiple factors, but cost reduction is likely among the benefits. “We want to get a model that is going to be more aligned to what we think is the right thing, is going to be more optimized to what we see users like in terms of the results we’re getting, and is going to be faster and cheaper for customers eventually than using the frontier models like Opus,” Shlomo said. As for Base44 itself, cost reduction isn’t as clear cut. In a press release, the company explained that “ownership of the model gives Base44 direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time.” Even with a delayed payoff, improved margins would be good news for Base44’s parent company, which recently announced it wouldlay off 20% of its workforce. In contrast, Base44 has been growing in headcount since the acquisition — and announced it hadpassed $100 million in annual recurring revenuea few months ago. That’s still less than Lovable, which said ithit $500 million in ARR earlier this month. But Shlomo is betting that the “huge engineering effort” to develop Base1 will cement Base44’s positioning as the “only vertically integrated vibe-coding application — meaning, in Userovici’s terms, a player that owns its distribution, data, and infrastructure all at once.
View
Gemini’s personalized AI image generation is now free for US users
Googleannouncedon Monday that the Gemini app is now offering its personalized Nano Banana-powered image generation feature to a broader audience. Starting today, all eligible users in the U.S. can access the feature for free, a service that waspreviously only availableto Plus, Pro, and Ultra subscribers. Google initially announced that Gemini’s Personal Intelligence feature would get Nano Banana-powered image generation back in April, allowing users to create images that reflect their unique interests. This means that images can be generated based on Gemini’s understanding of your likes and preferences without you having to specify them in your prompt. Gemini utilizes data from your Google account connections — such as Gmail, Google Photos, YouTube, and Search — to achieve this. For example, instead of saying, “Create an illustration of me and my favorite things, such as coffee and baking,” you can simply request, “Create an illustration of me and my favorite things.” Gemini can also pull actual images of you from Google Photos, so you don’t need to manually upload photos. Google initially rolled out the Personal Intelligence feature earlier this year, making itwidely availableto all U.S. users in March. The company recentlyexpandedthis functionality to users in India and Japan. Personal Intelligence is an opt-in feature, allowing you to decide which apps Gemini can access. Once enabled, it is set as the default for every prompt, but you can disable it using a new toggle in the Tools menu. Additionally, last month, Googleannouncedseveral upcoming updates for the Gemini app, including a new “Daily Brief” feature, a revamped interface, access to AI video model Gemini Omni, and a personal AI agent named Gemini Spark. Notably, Google’s AI chatbot Gemini surpassed 750 million monthly active users (MAUs) earlier this year, reinforcing its position as a major player in the AI space.
View
TIDAL cracks down on AI music by cutting off monetization
Music streaming serviceTIDALis the latest to take aim at AI-generated music with the introduction ofa new policythat will prevent fully AI-generated music from making money on its platform. In addition, TIDAL will use automated tools to remove AI-generated music that attempts to impersonate an artist or a group, the company said. “We are committed to protecting and rewarding organic creativity to avoid compromising an artist’s ability to connect with and build their fandom from TIDAL subscribers. Many have told us they do not want to be exposed to — or prompted to listen to — wholly AI-generated music,” wroteTonyGervino, TIDAL EVP and editor-in-chief, in anannouncement. He clarified that TIDAL’s new policy was not meant to “bash technological advancement,” but rather focuses on protecting and rewarding “organic creativity” from artists. With the changes, fully AI-generated music on TIDAL will be identified and tagged as such, allowing listeners to see an “AI” badge next to any tracks deemed to be 100% AI. These tunes will not be able to be monetized or collect royalties, and will not be eligible for direct-to-fan sales, the company noted. TIDAL’s policy joins others in the streaming music space, where services like Spotify, Apple Music, Deezer, andQobuzhave developed their own policies to address the growing number of AI-generated tracks filling their services. Spotify last yearrevamped its policies to label AI musicand better filter spam, while still acknowledging that AI tools would be used in the music-creation process to varying degrees. Apple Music alsotook the tagging approach. Deezer, which said that44% of all new musicuploaded to its platform daily is AI-generated, has taken a tougher position. It actively removes AI tracks from recommendations and excludes them from editorial playlists. It alsooffers its AI-detection technology to rivalsandprovides a consumer-facing toolthat lets you see if AI music has slipped into your playlists on competing services. TIDAL’s policy could be an interesting test to see if demonetization could be the thing to slow the deluge of AI music, which many listeners aren’t interested in. “Regardless of what you are reading elsewhere, AI’s takeover of the music industry (and your recommendations) isn’t inevitable if we take even greater steps now to monitor and control it,” noted Gervino. The company said the new policy is a “living document,” meaning it’s open to changes as the space evolves. It goes into effect on July 15, 2026.
View
Cursor now has a mobile app for guiding your coding agent on the go
Cursor isn’t lettingthe $60 billion SpaceX acquisitionslow it down. On Monday, the companyannounceda new mobile appfor iOSdevices designed for users who want to prompt coding agents directly from their phone. The app ties into the Cursor 2.0 changesunveiled in October, which shifted the service towards independent coding agents. With the mobile app, users can spin up new coding agents or interact with agents that were initiated from the desktop client. Cursor’s move to mobile follows similar apps from Anthropic and OpenAI, both of which offer ways to interact with their coding tools on mobile. It’s part of a broader shift in AI-based coding tools, which are increasingly abstracting away from written code and towards oversight of code-writing agents. With no need to access large code bases, many developers are switching away from multi-monitor desktop setups in favor of phones, which allow continuous conversations with remote agents. In a recent talk, Anthropic’s head of Claude Code, Boris Cherny, said he had almost entirely switched to mobile AI coding as a result. “Most of my coding now is on my phone,” Cherny said in the talk. “I would have said ‘you’re crazy’ if you told me that six months ago, but yeah, here we are.”
View
Arena, the AI leaderboard everyone uses, is now a $100M business
Just eight months after launching its commercial service, AI leaderboard providerArena, which originated as a research project at UC Berkeley in 2023, has reached $100 million in annualized run-rate revenue. Arena is best known for its popular crowdsourced AI model performance leaderboard, generated from over 10 million user evaluations. Its consumer website lets a user type a prompt it sends to two models; afterwards, the user chooses which model did a better job. While Arena’s popular AI model leaderboard is free for public use, the company began generating revenue from its platform in September when it introducedAI Evaluations, a service that provides model labs and enterprises with deep-dive performance analytics gathered from its community. Arena’s rapid revenue growth shows that its commercial offerings are as popular with customers as they are with its community of evaluators, who are frequently drawn to the platform for early access to the latest, often unreleased, AI models. “A lot of people don’t even understand that our business is making any money at all; people still see us as like an open-source project,” Anastasios Angelopoulos, Arena’s co-founder and CEO, told TechCrunch. While Arena calls its revenue milestone ARR, a term that traditionally stood forannualized recurring revenue, Angelopoulos clarified that the company charges customers for “consumption,” which means that its revenue is not recurring. While Arena doesn’t have direct competitors – Yupp, another crowdsourced AI model-picking startup,shut downin March— Angelopoulos said the company competes “for the same dollar” with human labeling startups like Mercor, Surge, and Scale AI, all of which assist model makers in refining their AI during post-training. As AI providers strive to maximize model performance, their appetite for post-training refinement services continues to surge. When Arena announced in January that it raised a $150 million Series A at a post-money valuation of $1.7 billion, its annualized revenue was$30 million. Elsewhere, Handshake’s gross annualized revenue from AI training has nearly doubled since January, climbing from $550 million to nearly $1 billion, The Informationreportedin April. Mercor’s annualized revenue also topped $1 billion earlier this year, up from $500 million last September,accordingto The Information. Arena ranks models on a variety of tasks such as text, coding, vision, and image generation, as well as complex, long-running workflows through its recently introduced Agent Mode. Along with Angelopoulos (pictured left), Arena was co-founded by fellow UC Berkeley postdoctoral student Wei-Lin Chiang (pictured center), who serves as the startup’s CTO. The startup was also co-founded by Ion Stoica (pictured right), the renowned UC Berkeley professor and Databricks co-founder who advised the project before it incorporated as a company in April 2025. Arena has raised a total of $250 million from investors including Felicis, Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, Laude Ventures, and UC Investments.
View
South Korean tech giants commit over $550B to ease ‘ RAMageddon’
The world’s two largest memory chip companies plan to invest $518 billion (~800 trillion won) to build four new memory fabs in southwestern South Korea, a region that has historically attracted little semiconductor investment. The announcement is part of the country’s sweeping national investment plan spanning semiconductors, AI data centers, and physical AI, which was unveiled at a presidential briefing on Monday, with the chairmen of Samsung and SK Hynix in attendance. The plan breaks down into three buckets. In the memory chip bucket is $518 billion for four new memory fabs in the southwest, plus $52 billion for an HBM (high bandwidth memory) packaging hub in the central region. Then there’s another $356 billion (550 trillion won) for AI data centers to be built by Korean tech and energy behemoths such as SK, GS and Naver through 2035. All told, South Korean tech companies have committed to spend over $900 billion on AI and the demands for chips it is creating. With this, the nation hopes to catapult itself into becoming more of an AI power player than it already is. Currently, Samsung and SK Hynix (along with U.S. memory chip maker Micron) are all enjoying record demand from what’s been calledRAMageddon, a worldwide shortage of memory chips caused by the AI buildout. “Semiconductors, physical AI, and AI data centers are the triple axis for South Korea’s next industrial era,” President Jae Myung Lee saidin a televised addressMonday, calling 2026 the year South Korea must establish itself as an “irreplaceable” industrial power. Lee said existing chip facilities in Yongin and Pyeongtaek, the heart of South Korea’s semiconductor belt just south of Seoul, have “already reached their limits,” and urged companies to accelerate investment in the southwest, hoping to spreading the AI wealth beyond the nation’s capital. “We must secure overwhelming production capacity in advance,” he said. Yet, Lee pushed back against media reports that the government had pressured companies into the investments,reportedly sayingthe decisions reflected the companies’ own judgment. “The government’s role is to invest its capabilities so that companies can invest without losses and with better prospects,” he was quoted as saying. Samsungseparately published a press release Monday, announcing plans to invest 2,655 trillion won (~$1.7 trillion) over the next decade, with 425 trillion won earmarked for the Honam region, the southwestern corner of the Korean peninsula. The company cited expected incentives around power, water, workforce, and living conditions as key factors in selecting Gwangju, roughly 300 kilometers south of Seoul, for a new semiconductor fab, alongside an AI data center in Haenam, at the southern tip of the peninsula. That is not an outlandish sum compared to U.S. tech giants Alphabet, Amazon, Meta, and Microsoft, who will collectively spend $650 billion on AI infrastructure this year alone,according to Reuters. Meanwhile, SK Group announced a 2,100 trillion won (~$1.4 trillion) medium-to-long term investment roadmap, 1,100 trillion won to expand semiconductor production capacity and 1,000 trillion won for AI data centers nationwide. SK Hynix, the group’s core semiconductor affiliate, is central to the chip expansion push, while SK Telecom will lead the buildout of 15 gigawatts of AI data center capacity across the country. Whether the ambition translates into execution is another question. Deep tech industries like semiconductors and AI don’t move on political or even customer demand timelines. Fabs take years to build and the risk is that, by the time they are ready, the demand that caused them will have ebbed, leaving companies with oversupply and crashing prices. For now, the world’s AI chip supply chain, especially those hungry for all things memory, will be watching to see if South Korea can pull it off.
View
Anthropic and Gov. Newsom forge deal allowing California government to use Claude at half price
Governor Gavin Newsom (D-CA) and Anthropic havemade a dealthat allows California government agencies to use Claude at a discounted price. This agreement comes at a time when businesses arestrugglingto manage the hefty costs of enterprise subscriptions to AI tools. Under the deal, all state agencies and local governments will have access to Claude, Anthropic’s AI chatbot, as well as training and support from Anthropic. A press release from the Governor’s office says that Claude will help state employees draft documents and analyze information. “AI should not replace the human work of government; it should help our workers move faster, solve problems more effectively, and deliver better results for Californians,” Governor Newsom said in astatement. This deal follows Newsom’s Marchexecutive orderthat intends to accelerate the use of AI “to make government more efficient” while also maintaining stronger safety standards. “While others in Washington are designing policy and creating contracts in the shadow of misuse, we’re focused on doing this the right way,” Newsomsaidat the time. As Anthropic forges a closer relationship with the state of California, the federal government has made an enemy out of the OpenAI rival. Earlier this year, Anthropic and the U.S. Department of Defenseclashedover a contract that would give the government agency permission to deploy Claude for any lawful use. Anthropic sought to explicitly carve out protections that prevent the government from using its technology to surveil Americans or deploy autonomous weapons without human oversight. But Defense Secretary Pete Hegseth refused, and the agency signed a deal with OpenAI instead. The government went as far as to declare Anthropic a “supply-chain risk,” preventing the company from working with any other Pentagon contractors. While the state’s path clearly diverges from the actions of the federal government, California’s CIO and Department of Technology director Chris Giventold POLITICOthat the supply-chain risk designation “just didn’t come up” while negotiating this Anthropic contract.
View
Tech Mahindra Partners With Perplexity to Strengthen AI-Led Sales Operations
Tech Mahindra will roll out Perplexity Enterprise Pro across its sales and customer-facing teams.
View
