Latest AI News
Mistral AI inks a deal with global consulting giant Accenture
Enterprises havestruggled to find a return on investmentfrom adopting AI tools. Now AI companies are trying a new tactic to get their tech to stick by partnering with consultants. On Thursday, French AI research lab Mistral AI and global consulting giant Accentureannounced a multiyear partnership. The deal entails the two companies working together to develop enterprise tech powered by Mistral’s AI models for clients. Financial terms and duration of the deal were not disclosed. TechCrunch reached out for more information. This deal also includes Accenture becoming a Mistral customer and rolling out its technology to its underlying employees. While Mistral is often seen as a smaller European peer to the sprawling U.S. AI “startups,” the partnership proves that Mistral can still land the same sizable customers, given that OpenAI and Anthropic recently announceddeals with Accentureas well. The agreement comes as AI companies are increasingly partnering with consultants. For instance, OpenAI just announced its“Frontier Alliance”initiative with four large consulting firms, including Accenture, on Monday to help push its newOpenAI FrontierAI agent governance platform to enterprise customers. Anthropic is also partnered withIBMandDeloitte. Will partnering with consulting firms be the trick to finding more enterprise AI adoption? It’s unclear, but what is clear is that AI companies are definitely trying to find out.
View
Sophia Space raises $10M seed to demo novel space computers
As space companies itch to push the most advanced chips into orbit, the problem of cooling those high-powered processors is top of mind. “It’s cold in space … [but] there’s no airflow, and so the only way to dissipate is through conduction,” Nvidia CEO Jensen Huang said when asked about space-based data centers during his firm’s most recent earnings call. Now, Sophia Space has raised $10 million from investors, including Alpha Funds, KDDI Green Partners Fund, and Unlock Venture Partners. The company plans to prove out a new approach to passively cooling space computers on the ground, then buy a satellite bus from Apex Space and show that it works in orbit by late 2027 or early 2028. Companies like SpaceX, Google, or Starcloud are examining traditional satellite form factors for their proposed space data center constellations, which rely on large radiators to keep chips in optimal thermal condition. But Sophia Space’s founders — CTO Leon Alkalai, CEO Rob DeMillo, and chief growth officer Brian Monnin — have a different approach. The company’s tech comes from an unusual source: a $100-million-endowed program at Caltech to develop orbital solar plants that would beam electricity to Earth below. The researchers ultimately settled on a sail-like structure that is thin and flexible compared to boxy, traditional satellites. While technical and regulatory challenges make producing electricity for Earth difficult, Alkalai, a fellow at the Caltech-managed Jet Propulsion Laboratory, was struck by the idea of using the design to power space-based processors. (Aetherflux, a space solar power startup, has had asimilar realization.) Sophia, an Nvidia partner, has designed modular server racks with integrated solar panels it calls TILES, which are 1 meter by 1 meter in area and a few centimeters in depth. By adopting this thin form factor, DeMillo says that processors can sit against a passive heat spreader, eliminating the need for active cooling. He expects 92% of the power it generates will go to processing, a significant gain on traditional designs. This design requires, however, a sophisticated software management system to balance activity across the processors. By the 2030s, Sophia hopes to be building larger space data centers out of thousands of TILEs, envisioning a 50-meter-by-50-meter structure delivering 1 MW of computing power. DeMillo argues that attempting to build space data centers with less efficient systems will not be economical and that a single structure rather than a distributed network linked by lasers will be easier to execute. First, however, Sophia plans to begin by offering its TILEs to satellite operators that require compute solutions on orbit. Potential partners include Earth-observation satellites collecting large amounts of sensor data, missile warning and tracking systems that the Pentagon is investing billions of dollars to build, or even increasingly complex communications networks. “The dirty little secret of the satellite industry is we’ve got all these amazing sensors up there that produce terabytes, or even petabytes, of data every few minutes, and they throw most of it out because they can’t do the computing on board and they can’t get round trip back and forth to the surface fast enough,” DeMillo told TechCrunch.
View
So, we’re getting Prada Meta AI glasses, right?
Could Meta be preparing to launch a Prada version of its Meta AI glasses? That’s the speculation after Mark Zuckerberg and his wife, Priscilla, were spotted sitting in the front row of Prada’s Fall/Winter 2026 Fashion Week show in Milan on Thursday. The social media exec was seenchatting with his seatmate, Lorenzo Bertelli, Prada’s chief merchandising officer and son of head designer Miuccia Prada. While Zuckerberg has been working to polish his image in recent years, includingwith upgraded threads, it’s likely that he wasn’t at Prada for the fashion, but rather because of an upcoming collaboration with the brand. CNBC reportedlast summer thatPrada AI glasses were in the works, among others. However, Meta has yet to publicly announce such a deal. (The company has not yet responded to a request for comment about Zuckerberg’s presence in Milan.) Surprise guests at Pradapic.twitter.com/BPHqLo9oaW EssilorLuxottica, the French-Italian eyewear brand and Ray-Ban maker, has been working with Meta on these high-tech devices since their debut, initially under the Ray-Ban Stories brand. This month, the company announced itsold over 7 million AI glasses in 2025, up from 2 million in the prior year. Those sales included both Ray-Ban Meta andOakley Metaglasses, the latter designed more for the athletic types. Now, it seems, Prada AI glasses could be next, given that Prada and EssilorLuxottica already renewed their licensing deal for eyewear under the Prada and Miu Miu brands for the nextten years. (The existing agreement, which expired on December 31, 2025, was later extended through December 31, 2030, with the provision for renewal until December 31, 2035.) Prada AI glasses could give Meta a foothold in the high-fashion market, a niche that its Oakleys and Ray-Bans don’t yet fill. Establishing the glasses as a luxury symbol could also benefit Meta’s brand overall. However, there are some concerns that AI glasses aren’t the right fit for a world that’s seeing an increased consumer backlash against surveillance devices, which have recently led people torip outtheirRing doorbellsandsmash Flock cameras. This shift could see Meta reconsidering whether it willadd facial-recognitionfeatures to its glasses, asThe New York Timesrecently reported. The news drew criticism for what had otherwise been a modestly successful tech product and has even promptedone developer to build an appthat will warn you if someone is wearing the AI glasses near you.
View
Google launches Nano Banana 2 model with faster image generation
Google today announced the latest version of itspopular image generation model, Nano Banana 2. The new model, which is technically Gemini 3.1 Flash Image, can create more realistic images than its predecessor. The model will also now become the default in the Gemini app for its Fast, Thinking, and Pro modes. The company first releasedNano Banana in August 2025, prompting people to generate millions of images in the Gemini app, especially incountries like India. In November, the company releasedNano Banana Pro, which allows users to create more detailed and high-quality images. The new Nano Banana 2 retains some of the high-fidelity characteristics of the Pro model but produces images faster. The company says you can create images with a resolution ranging from 512px to 4K, in different aspect ratios. Nano Banana 2 can maintain character consistency for up to five characters and fidelity of up to 14 objects in one workflow for better storytelling. Users can also issue complex requests with detailed nuances for image generation, Google says. In addition, users can create media with more vibrant lighting, richer textures, and sharper detail. With the launch, Nano Banana 2 will become the default model for image generation across all apps in the Gemini app. The company is also making it the default model for image generation in its video editing tool,Flow. In Search, Nano Banana 2 will become the default for Google Search results via Google Lens and in AI Mode across 141 countries on the Google app and on the web across desktop and mobile. OnGoogle’s higher-end plans, Google AI Pro and Ultra, subscribers can continue to use Nano Banana Pro for specialized tasks by regenerating images via the three-dot menu. For developers, Nano Banana 2 will be available in preview through the Gemini API, Gemini CLI, and the Vertex API. It will also be available through AI Studio andthe company’s development tool Antigravity, which was released last November. The company said that all images created through the new model will have a SynthID watermark, which is Google’s mark to denote AI-generated images. The images are also interoperable withC2PA Content Credentials, created by an industry body consisting of companies like Adobe, Microsoft, Google, OpenAI, and Meta. Google said that since launching the SynthID verification inthe Gemini app in November, people have used it over 20 million times.
View
Bumble adds AI-powered photo feedback and profile guidance tools
Bumbleannouncedon Thursday that it’s adding a series of AI-driven features intended to help turn matches into lasting connections, including those that offer feedback and guidance on users’ bios, photos, and prompts. The dating app’s new AI-suggested profile guidance tool will roll out globally and give “personalized, actionable feedback” on users’ bios and prompts. For users in the U.S., the profile guidance feature can be augmented with an AI photo feedback tool, which can “help you choose the best photos and show up as your most authentic self.” According to Bumble’s blog post explaining these features, it doesn’t seem like the insights from these AI tools are particularly groundbreaking — for example, Bumble says that its AI photo tool might encourage you to ditch photos where you’re wearing sunglasses that cover your face, and add a wider variety of photos, like ones taken outdoors or with friends. It’s advice you could’ve easily gotten from a friend 10 years ago, but it’s still new information to many users. In Canada, Bumble is testing another, non-AI feature called “Suggest a Date.” When a conversation stalls, a user can signal that they are open to meeting in person, which the company says is “a simple way to signal that they’re ready to connect offline.” Of course, another way for people to “signal that they’re ready to connect offline” is to literally ask someone on a date. But realistically, it doesn’t seem like users are taking the plunge, so having an in-app way to indicate interest may motivate some potential couples to move their conversation IRL. “With Suggest a Date, we’re creating a clear expression of intent and giving members a way to bypass the traditional back-and-forth and move toward meeting in real life,” Bumble CTO Vivek Sagi said in a statement. “When we reduce friction at the moments that matter most, we help people connect with clarity and confidence, and increase the likelihood of meaningful relationships forming offline.” Bumble and other popular dating apps, like Match Group’s Tinder and Hinge, have all embraced AI-powered features in recent months. For instance, in December, Hinge introduced a tool to help generate more interesting conversation starters than “How are you?” Tinder may take things a step further. In Australia, Tinder is piloting a tool calledChemistry, which asks users to provide the app with access to their camera roll, which is a concerning amount of data to feed into an AI tool. Based on a user’s camera roll and answers to a series of questions, the AI can learn more about someone’s interests and personality to supposedly reduce “swipe fatigue” and suggest better matches. Meta’s Facebook Dating tool does something similar — in October, it launched a feature thatasks to use its AI on photos in your camera rollthat you haven’t yet shared in order to suggest AI edits. As these companies try to come up with new ways to keep users happy, some young people have thrown in the towel on online dating altogether, instead seeking morereal-world experiencesthat are not intermediated by an app.
View
Read AI launches an email-based ‘digital twin’ to help you with schedules and answers
Meeting notetaker Read AI on Thursday launched an AI-powered email-based assistant called Ada, saying it helps users manage their schedules, answer questions based on a company’s knowledge base, and reply to out-of-office emails. The company is calling Ada a “digital twin” that handles tasks for you around the clock. Read AI said that the assistant will be available to all users, and they can start configuring it by sending an email to “[email protected]” and writing “Get me started.” When you ask Ada to find a time to meet with someone, it replies to the other person in the thread with your availability. If the other person replies that they are unavailable at those times and would like a different time slot, Ada responds with new options. While Ada has access to your calendar through Read AI, it does not reveal the nature of those meetings with other people. Ada can also answer questions using a company’s knowledge base, topics discussed in your prior meetings, and public internet searches. For instance, you can ask, “Ada, can you provide an update on how we are tracking for Q1 goals?” to get information. If someone else asks a question in a thread, Ada will prepare a response for you and help you refine it before it is sent to the other person. The startup said that Ada doesn’t reveal any sensitive information without your permission. Read AI’s VP of Product, Justin Farris, said that the new feature doesn’t rely on MCPs (model context protocols, a technical standard for connecting AI tools to external services), and instead builds a knowledge graph based on meeting data and connected services for more contextual answers. He added that over time, the assistant will also take proactive actions for you. For instance, if you mentioned a follow-up item in a meeting, Ada will ask you to set that up after the meeting with contextual data. “The way I describe our solution is that when you are bringing on a new employee, you train them. When you add Ada to your workflow and connect more services to give more context, it starts to ramp up and handle more tasks for you,” CEO David Shim told TechCrunch. The company said that while Ada currently works via email, it will soon be available on Slack and Teams. On the sidelines of Web Summit Qatar earlier this month, Shim told TechCrunch that the company now has over 5 million monthly active users and plans to grow that number to 10 million. He mentioned that the company sees 50,000 sign-ups every day and has a broader base of 100,000 users who consume Read AI’s content, like meeting summaries, without creating an account. For Read AI, the U.S. remains the largest market with strong international growth. While 60% of users are outside the U.S., the revenue is split roughly equally. The company, which has raised over$81 million in funding, is increasingly adding AI-powered tools to its suite. Last year, it launchedSearch Copilotfor knowledge discovery for users, and last month it added the ability to update customer-service relationship software, send custom emails from within a meeting report, and stay up to date on topics basedon internal and web knowledge. Other meeting notetakers are also offering new tools to extract more insights and actions from meeting notes. Last September, Granola added“recipes” in the form of repeatable promptsto surface knowledge from meeting data.Quill, which came out of stealth with a $6.5 million funding round this week, also connects to various tools like Linear, Notion, and CRMs, and aims to automate tasks.
View
Figma partners with OpenAI to bake in support for Codex
Figma is integrating OpenAI’s AI coding tool, Codex, to let users create and tweak designs from within their coding environments. The move comes a week after the design companystrucka similar partnership with Anthropic to integrate Claude Code. This integration lets users start working with a design in Figma or code in Codex, and move between the two platforms easily using Figma’s MCP (Model Context Protocol) server. Previously, users could bring details from Figma design files, Figma Make, or FigJam into Codex for code-based implementation. “With this integration, teams can build on their best ideas—not just their first idea—by combining the best of code with the creativity, collaboration, and craft that comes with Figma’s infinite canvas,” Loredan Crisan, Figma’s chief design officer, said. Loading the player… “The integration makes Codex powerful for a much broader range of builders and businesses because it doesn’t assume you’re ‘a designer’ or ‘an engineer’ first. Engineers can iterate visually without leaving their flow, and designers can work closer to real implementation without becoming full-time coders,” Codex product lead Alexander Embiricos said in a statement. OpenAI first launchedCodex as a command-line coding assistantlast year to compete with Anthropic’s much lauded Claude Code. Later, the company builtthe coding tool into ChatGPT, and earlier this month launcheda dedicated MacOS app for Codex. The MacOS app was downloadeda million times within the first weekof release. The company also releasedtwonewCodex models just days afterwards. OpenAI said that overa million users are using Codex weekly. Figma has been a prominent partner for OpenAI, and it wasone of the first companies to launch an appin ChatGPT in October 2025.
View
Trace raises $3M to solve the AI agent adoption problem in enterprise
For all their potential, AI agents have been slow to make an impact in the enterprise, and one new startup is betting that the reason they haven’t is a lack of context. Launched as part of Y Combinator’s 2025 summer cohort,Traceis a workflow orchestration startup aimed at filling that gap. The company maps complex corporate environments and processes so that agents have the context they need to scale quickly. “OpenAI and Anthropic are building these brilliant interns that can be leveraged within the company,” says Trace CEO Tim Cherkasov, referring to the AI labs’ tools. “We’re building the manager that knows where to put them.” On Thursday, the London-based company said it had raised $3 million in seed funding from Y Combinator, Zeno Ventures, Transpose Platform Management, Goodwater Capital, Formosa Capital, and WeFunder. Angel investors Benjamin Bryant and Kevin Moore also invested. Trace’s system starts by building a knowledge graph from a company’s existing tools — systems like email, Slack, and Airtable that shape the day-to-day working life of the firm. With that context in place, users can prompt the system with a high-level task — like “We need to design a new microsite” or “Lets develop our 2027 sales plan” — and Trace will come back with a step-by-step workflow, delegating some tasks to AI agents and assigning others to human workers. When the system does invoke an AI agent, it will prompt it with the specific data needed to complete its sub-task. The idea is to automate away the delicate work of on-boarding AI agents, one of the biggest blockers for actual deployment within companies. With so many companies focused on agentic AI, Trace will have plenty of competition.Earlier this week, Anthropic launched its own take on enterprise agents, focused on pre-built plugins for specific departmental functions. And many of the workplace productivity services Trace will be drawing from, likeAtlassian’s Jira, are launching their own agents, which will potentially compete with the startup’s system. But Trace’s founders believe their knowledge-graph approach will be the key to success, as they can build context engineering deep into the structure of agentic deployment. “2024 and 2025 was still about prompt engineering. Now we’ve moved from prompt engineering to context engineering,” says CTO Artur Romanov. “Whoever provides the best context at the right time is going to be the infrastructure on top of which the AI-first companies will be built. And we hope to be that infrastructure.”
View
Exhibit in Boston’s startup ecosystem at TechCrunch Founder Summit 2026
On June 9, 1,000+ founders, investors, and decision-makers will gather forTechCrunch Founder Summit 2026. This isn’t just foot traffic. It’s a full day of concentrated deal flow. An exhibit table turns that energy into measurable growth. You don’t need more awareness. You need customers, capital, and strategic partners. Secure your exhibit tableand put your startup at the center of the startup universe. Direct access to buyers:The Expo Hall is where founders and operators actively source tools and partners. Capture leads instantly through the event app and start building your pipeline on the spot. Where handshakes replace cold pitches:Move from cold outreach to face-to-face conversations. With access to roundtables, breakouts, and curated networking, your team can turn visibility into velocity. Credibility that compounds:Your startup is featured across the Founder Summit website, app, sponsor directory, announcements, and closing ceremony — reinforcing trust with customers and investors alike. Multi-channel ROI:Lead gen tools, sponsor listings, and team-wide passes mean you’re not just exhibiting. You’re executing across sales, marketing, and fundraising at the same time. Deploy your team strategically:With five total passes, you can split focus across sales, partnerships, hiring, and curated meetings to maximize every hour onsite. Spots are limited and first-come, first-served. If you’re ready to turn three days into months of pipeline,reserve your exhibit table nowbefore your competitor does.
View
2 days left: Lock in the best discounts for TechCrunch Disrupt 2026
Super Early Bird pricing endstomorrow, February 27, at 11:59 p.m. PT. After that, prices forTechCrunch Disrupt 2026go up. Miss this, and you’ll be paying more for the same access to one of the most anticipated tech events of the year.Register nowto secure discounts of up to $680 on your pass, or up to 30% ongroup passes. If you want to raise capital, hire top talent, launch your startup, or discover your next portfolio company, don’t missDisrupt, taking place October 13–15 at San Francisco’s Moscone West. Here’s what you’ll gain by attending: Founder Pass: Accelerate growth with the right insights, tools, and connections. Meet investors aligned with your startup. Investor Pass: Discover standout startups and expand your portfolio with curated access. Use matchmaking tools to make every conversation count. This window to the lowest ticket rates of the year is closing after tomorrow ends.Register nowto secure your ticket with up to a $680 discount. Or save up to 30% withcommunity passesof 4+.
View
Indian IT Builds the Missing Layers in Open-Source Agentic AI Frameworks
Tech Mahindra and Persistent detail enterprise safeguards as study flags immature frameworks.
View
Google Unveils 10-Hour AI Certificate in Latest Learning Push
The curriculum focuses on applied skills across research, strategic planning, professional communication, content creation, data analysis and workflow automation.
View
