Latest AI News
Sam Altman is ready to decelerate
OpenAI CEO Sam Altman says that it may be time to “pace” AI development so the world will be ready for it. “We may have to pace the rate of AI development to give ourselves enough time for society to harden around some of these new capability levels,” hetoldPatrick O’Shaughnessy, the host of the Invest Like the Best podcast, while also “trying to figure out how we do that in a way that does not feel like regulatory capture for anyone and also does not feel like collusion among the frontier labs.” BothOpenAIandAnthropiccame out in support ofa petitioncirculated by employees at the frontier labs, calling on the US government to “support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.” Altman has avoided signing on to past campaigns to slow AI progress,callinga 2023 open letter that proposed a similar slowdown “missing most technical nuance about where we need the pause.” Apparently, he’s softened on the idea, thanks in part tothe incidentwhere one of OpenAI’s advanced models managed to break out of a secure computing environment and hack into Hugging Face, an online model database, using several zero-day exploits. On the podcast, the OpenAI co-founder called it an “extremely sci-fi cyber incident…[t]his is the first security incident that I have felt very viscerally.” OpenAI researchers have paused training on that model while they work out how to keep their sandbox secure. But Altman said that as models become more powerful, the need to “pace” their development could become key to safe deployment. While model safety and alignment has always been a concern in the AI world, the arrival of Anthropic’s highly capable Mythos model earlier this year has turned hypotheticals into real-world problems. However, the industry has a trust problem, and challenging economics that give major players an incentive to play up the dangers of their systems in ways that experts disagree about — for example, whether or not Anthropic’s Fable model should have been briefly banned from use or not. Similarly, when Kimi K3, a large, open-weight model built in China, was released, OpenAI head of strategic futuresDean W. Ball said that it threatened the economics of frontier labs, making it difficult to separate their safety concerns from their financial interest. “I think a lot of the talk about safety concerns is well-founded, and then a lot of it is about people that just really, even if it’s slightly subconscious, want to concentrate power,” Altman mused, in what reads as a dig at his rival, Anthropic CEO Dario Amodei, who signed the petition. “I am terrified of a world where the very real fears of AI are used as a way to say, ‘Only this small group of people can have it because it’s too dangerous, and only they understand it, but don’t worry, like, they’re gonna make the right decisions for all of us.’ I don’t believe in that.” Still, OpenAI has pushed back against efforts to develop government rules for AI models, instead preferring an industry-led approach where AI labs would create ostensibly independent organizations that would evaluate the security of models and the safety approach of their makers.The challenge for both that approach to regulation and any efforts to slow AI development will be getting the various players in the industry on the same page, whether they are rival frontier labs in the U.S. or competitors in China. This story was updated to include OpenAI and Anthropic’s support of the “Pacing the Frontier” petition.
View
MCP startup Runlayer accuses Rippling of stealing its product idea
Runlayer, a startup that offers a secure Model Context Protocol gateway — a standard for letting AI models and agents securely pull in outside data and tools — has filed a lawsuit against HR software startup Rippling, according to the complaint seen by TechCrunch. The lawsuit is a cautionary tale for anyone selling AI infrastructure to enterprise customers, especially to other tech companies, that increasingly have the engineering muscle to just build the thing themselves. In the suit, Runlayer describes an extensive product trial conducted by Rippling as a prospective customer, during which the MCP startup shared everything from its product roadmap to its actual source code. The parties signed a mutual non-disclosure agreement and Rippling signed a product trial agreement with a clause that forbade it from copying Runlayer’s intellectual property or making derivative works, which is standard boilerplate in enterprise software trials. Runlayer says in the complaint that Rippling’s evaluation involved “nearly a year of intensive engineering collaboration.” But in the end, the two could not agree on a price, so Runlayer ended the product trial. Shortly after that, Runlayer alleges that a “Rippling insider” texted Runlayer founder and CEO Andrew Berman to inform him of “a project internally to build essentially a clone o[f] Runlayer … it’s almost a 1 to 1 copy of Runlayer.” Runlayer claims in the suit that Rippling’s product must have been based on the startup’s intellectual property and therefore constitutes trade secret misappropriation, unfair competition, and breach of contract. Rippling has confirmed to TechCrunch that it is indeed launching its own MCP gateway, though a spokesperson denies Runlayer’s allegations about misusing its IP. “Runlayer’s panicked effort to avoid competition by fabricating claims is not an effective way to deal with its business failures. Rippling is launching a superior product for connecting AI tools to business data using only our proprietary information — we have every reason to win in this market,” a Rippling spokesperson tells TechCrunch. Runlayer has retained white-shoe law firm Sullivan & Cromwell. That doesn’t mean Runlayer will, or even should, win this suit, but the same way a marquee VC lends a startup some credibility, a marquee law firm lends a lawsuit some credibility, at least optically. The more interesting part about this suit is really the inside peek it provides at the trials and tribulations of selling complex AI infrastructure into the enterprise, particularly to other tech companies. Enterprise sales notoriously take a long time to close, often because they hinge on this kind of deep, hands-on trial. MCP gateways in particular are getting crowded.Anthropic launched MCP as an open source protocolin November 2024. It’s now one of the basic building blocks of AI interoperability, giving models and agents a secure way to access external data sources and services. MCP gateway products add control, security, and other features, especially for managing agents, and the field has grown considerably more competitive since Runlayerlaunched its productin the middle of last year and raised a total of$42 million, including from Khosla Ventures and Felicis. Even after an intense trial, an enterprise may simply opt to build the tool in-house. Both sides are stuck between a rock and a hard place.
View
Bot-detection startup Spur nabs $200M from Insight
Spur Intelligence, a cybersecurity startup based in Lake Mary, Florida, has raised a $200 million round led by Insight Partners. Spur, founded by two former Defense Department engineers in 2017 — five years before ChatGPT’s public launch — was prescient. The startup’s tech helps enterprises distinguish legitimate human users from increasingly well-hidden bot traffic to help identify fake users and threats. “As sophisticated criminal VPNs, residential proxy networks, and anonymization infrastructure proliferate, organizations are increasingly operating with a critical blind spot: they can see the activity, but not the infrastructure behind it,” Insight’s Thomas Krane said in a written statement. Detecting malicious traffic has, of course, been a hill corporate security teams have been climbing for eons. But nothing compares to the onslaught facing them today. As of mid-2026, bots are nowmore activeon the internet than humans, Cloudflare reported last month. “Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet’s history,” Cloudflare founder and CEO Matthew Prince postedon X last month, pointing to his company’s latest traffic report.
View
Data centers may face temporary power cuts to prevent blackouts on largest US grid
The largest electrical grid in the U.S. has struggled to cope with an onslaught of data centers. Now, after an auction to add more generating capacity fell short, the grid’s operator, PJM Interconnection, has said it will cut off data centers and other large users during power shortages. The decision arrives as the breakneck pace of data center construction has grid operators scrambling to generate power. By 2035, data centers are expected to use4x more electricitythan they do today. PJM won’t start curtailing supply until June 2027, and the cuts will only apply to data centers that are 50 megawatts or larger. The grid operator is running another auction for new generating capacity. Similar to other demand response programs, which have existed for decades and typically include large users like manufacturers, the customers who have their power cut will be compensated. Such programs typically give customers advance notice, ranging from 30 minutes to a few days, depending on forecasted demand. The move will likely spur many new data centers — and potentially existing ones — to set up their own sources of on-site power. Those that don’t will probably rely on backup generators, which tend to be costlier to run and frequently more polluting. Many data centersfavor diesel generatorssince the fuel is widely available and can be stored on-site.Federal regulationsallow such generators to be used for up to 50 hours per year for demand response events, and up to 100 hours per year for events like emergencies and maintenance. This week, Vantage Data Centers came under fire for itsapparent coordinationwith Virginia environmental regulators to cast doubt on a report that said diesel backup generators could contribute totens of millions of dollarsin annual health damages for people living near a96 megawatt data centerin Northern Virginia. PJM hascome under firein recent months for the way it has managed new generating capacity and large new users, including data centers. The grid operator’s territory runs from Virginia to Illinois, covering 67 million customers. Over the last year, wholesale electricity prices havenearly doubled, and PJM’s independent market monitor blamed data centers for much of the increase.
View
Exclusive: Samsung's JB Park on Galaxy AI Costs, Monetisation, and What Stays Free
For years, the smartphone industry has operated on a principle where consumers paid upfront for the latest hardware, and the software benefits were bundled at no extra cost. Every camera upgrade, display enhancement, and new-gen processor factored into the retail price of the overall device. However, the dawn of generative AI (short for artificial intelligence) has fundamentally disrupted this economic principle for the industry.
View
Claude Artifacts Found in Google Search, Raising Fresh Privacy Concerns
Claude users are facing renewed privacy concerns after publicly shared Artifacts created with the AI chatbot were found through Google Search, raising questions about how the platform handles content shared using public links. Reports indicate that while shared conversations have largely disappeared from Google's search results, some publicly shared apps, documents, spreadsheets, dashboards, and other AI-generated content remain searchable. The development has drawn attention to how publicly shared AI content can become more widely accessible than some users may expect.
View
Recursive Superintelligence signs $410 compute deal with Amazon
On Tuesday, the AI companyRecursive Superintelligenceannounced a $400 million compute deal with Amazon Web Services. The company, which emerged from stealth in May with $650 million in funding, is focused on building open-ended self-improving systems, a potentially compute-intensive approach to AI research. This multi-year deal is meant to provide flexibility as the company looks to scale up those systems. Recursive’s $410 million outlay represents the bulk of the company’s fundraising to date— but on a call with TechCrunch, Socher emphasized that he expected it to be the first of many such deals. Today’s announcement is “likely going to be one of the smallest compute deals we’re going to sign in the next few years,” Socher said. Recursive’s emphasis on self-improving AI systems means much of the budget that would traditionally go towards headcount and operations is put straight into compute, as the company seeks to automate its own product development process. “For us, it’s less about headcount and more about agent count,” Socher said. There’s no investment component to Amazon’s involvement, in contrast to major labs’ habit of hybrid investment arrangements. But the sheer scale of the commitment allows AWS to commit significant resources to supporting Recursive’s unique needs, which may help to draw in other foundation-level AI companies going forward. “Part of the agreement is that we’re going to co-develop infrastructure purpose-built for these types of company,” said Jason Bennett, VP for startups and venture capital at AWS. Recursive self-improvement (RSI) has long been seen as an inflection point for AI, with some expecting an explosion of progress once AI can be improved without human involvement. But as more labs and companies pursue the idea,the specific requirements have become ambiguous, with some predicting an imminent breakthrough while others characterize self-improvement as more of a continuum. But in Recursive’s case, the goal is to use the powers of RSI to develop actual products — and Socher expects to be releasing the earliest examples before the end of the year. “We are excited to build like really amazing products that people can use, and you will see those within a few months, not within a few quarters or years,” Socher says. “In October or so, you’ll see some actually tangible, useful things that you’ll be able to play around with.”
View
Fish Audio raises $52M seed to build AI voice models for creators and enterprises
The market for AI-generated voice models is massive. Creative use cases require AI voice models to be more expressive, while enterprises looking to automate customer support and sales ops need them to be more steerable. Palo Alto-based Fish Audio wants to cater to all of those use cases with its library of more than 15,000 natural language controls. Since launching last year, the startup now has more than 8 million people using the open-source or hosted versions of its models, and generates annual recurring revenue of $21 million. To continue building on that traction, the startup on Tuesday said it has raised $52 million in a seed round that was led by Coreline Ventures and Capital Today. The funding also saw participation from 359 Capital, Parable, Play Time, Alphalist Partners, Bayhouse Ventures, Carya Venture Partners, and HF0. Fish Audio started as a small project by former NVIDIA researcher Shijia Liao, who, frustrated by non-expressive synthetic voices available on the market, trained a voice generation model on a single GPU, which he then open-sourced. The Fish Speech repository on GitHub now has more than 31,000 stars and is used by indie developers, video game designers, and creators. The company has launched five models in the last year: four speech generation models and one speech-to-text model. It has open-sourced three of its speech generation models, but its latest S2.1 Pro model is available only through its paid API. Fish Audio offers paid monthly plans suited for creators and teams that unlock a set number of minutes of generation plus voice cloning features. The company also offers an enterprise version of its APIs and platform, and says organizations like HeyGen and Sanas are already using it. “Every enterprise has different use cases and different preferences. For example, companies like HeyGen, which use our voices to power AI avatars, want realism in voices; a gaming studio would want expressive voices for their characters; and voice agent companies like LiveKit want more natural-sounding and low-latency voices that are expressive enough for calls,” Cao said. One way the startup has built its library of voices is by asking users to submit their own voices for training its models, and compensating them if their voices are used. That resulted in some trouble a few months ago, however, assomecreatorsallegedthat their voices were uploaded to Fish Audio without their consent. The startup had a DMCA takedown process in place to address such concerns, but the takedowns themselves took a long time. Fish Audio’s CEO and co-founder Rissa Cao told TechCrunch that the company has now automated the takedown process. Creators can submit a short voice sample or a contract to prove that an uploaded voice belongs to them, and their voice will be taken off the startup’s platform in less than three minutes, she said. Still, that doesn’t prevent anyone from uploading an artist’s voice without their knowledge. And until the artist finds out, their voice will continue to be used on the platform unless they file for its removal. Osuke Honda, a partner at Coreline Ventures, said a community-driven model only works when creators trust the platform. “A community-centric approach can only become a durable advantage if creators trust the platform. That means consent, transparency, and attribution must be built into the product rather than treated as afterthoughts. I believe the industry needs to move toward verified voice ownership, clear licensing terms, easy reporting and takedown processes, and eventually revenue-sharing models where creators benefit financially when their voices are licensed or used commercially,” he said. Cao said when the startup was only offering its product as an open-source project with plans for creators, it was running efficiently and didn’t need outside capital. But it wanted to develop more advanced models, and also wanted to accommodate enterprises as investor interest was ramping up, which led it to seek capital. Looking ahead, Fish Audio plans to release an audio understanding model this year. It’s also building a speech-to-speech model. The speech generation market is crowded, with companies likeElevenLabs,WellSaid, Cartesia, Speechify,Async (previously Podcastle), andKrispcompeting for creators and enterprises’ budgets. According to Rico Mallozzi, a partner at 359 Capital, fine-grained controls for developers and cost-efficient model training will help Fish Audio compete better with big AI labs. “I think what they’ve been able to build, state-of-the-art models, with the team they have, compared to some of these other well-funded AI labs or companies, is incredible. It shows their technical acumen in closing the gap between artificial-sounding and human-like voices,” Mallozzi told TechCrunch over a call. The story has been updated to reflect that the company raised $52 million in the seed round.
View
The Forward-Deployed AI Engineer: The Multi-Hat Career Built for Enterprise AI
Building an AI prototype is easier than ever. Turning it into a secure, reliable, and adopted enterprise solution is still the hardest part.
View
Can India’s AI Data Centre Boom Keep Pace with the Tropics?
As AI infrastructure scales across India, industry leaders warn that imported technologies alone are not enough. The next breakthrough will depend on homegrown R&D tailored to India’s climate.
View
Why Japan May Be India’s Most Important Semiconductor Partner Yet
“No country is 100% self-sufficient, and it will not happen because of the complexity.”
View
How AMD Won Over OpenAI, Meta & Anthropic
AMD’s GPU always made the noise on paper, but in 2026, the company finally got the ecosystem endorsement it has long yearned for.
View
