Latest AI News
Elon Musk’s lawsuit is putting OpenAI’s safety record under the microscope
Elon Musk’s legal effort to dismantle OpenAI may hinge on how its for-profit subsidiary enhances or detracts from the frontier lab’s founding mission of ensuring that humanity benefits from artificial general intelligence. On Thursday, a federal court in Oakland, California, heard a former employee and board member say the company’s efforts to push AI products into the marketplace compromised its commitment to AI safety. Rosie Campbell joined the company’s AGI readiness team in 2021, and sheleft OpenAIin 2024 after her team was disbanded. Another safety-focused team, the Super Alignment team, was shut down in the same time period. “When I joined, it was very research-focused and common for people to talk about AGI and safety issues,” she testified. “Over time it became more like a product-focused organization.” Under cross-examination, Campbell acknowledged that significant funding was likely necessary for the lab’s goal of building AGI but said creating a super-intelligent computer model without the right safety measures in place wouldn’t fit with the mission of the organization she originally joined. Campbell pointed to an incident where Microsoft deployed a version of the company’s GPT-4 model in India through its Bing search engine before the model had been evaluated by the company’s Deployment Safety Board (DSB). The model itself did not present a huge risk, she said, but the company needed “to set strong precedents as the technology gets more powerful. We want to have good safety processes in place we know are being followed reliably.” OpenAI’s attorneys also had Campbell admit that in her “speculative opinion,” OpenAI’s safety approach is superior to that at xAI, the AI company that Musk founded that was acquired by SpaceX earlier this year. OpenAI releases evaluations of its models and sharesa safety frameworkpublicly, but the company declined to comment on its current approach to AGI alignment. Dylan Scandinaro, its current head of preparedness, was hired from Anthropic in February. Altmansaidthe hire would let him “sleep better tonight.” The deployment of GPT-4 in India, however, was one of the red flags that led OpenAI’s non-profit board to briefly fire CEO Sam Altman in 2023. That incident took place after employees, including then-chief scientist Ilya Sutskever and then-CTO Mira Murati, complained about Altman’s conflict-averse management style. Tasha McCauley, a member of the board at the time, testified about concerns that Altman was not forthcoming enough with the board for its unusual structure to function. McCauley also discussed awidely reportedpatternof Altman misleading the board. Notably, Altman lied to another board member about McCauley’s intention to remove Helen Toner, a third board member who published a white paper that included some implied criticism of OpenAI’s safety policy. Altman also failed to inform the board about the decision to launch ChatGPT publicly, and members were concerned about his lack of disclosure of potential conflicts of interest. “We are a non-profit board and our mandate was to be able to oversee the for-profit underneath us,” McCauley told the court. “Our primary way to do that was being called into question. We did not have a high degree of confidence at all to trust that the information being conveyed to us allowed us to make decisions in an informed way.” However, the decision to boot Altman came at the same time as a tender offer to the company’s employees. McCauley said that when OpenAI’s staff started to side with Altman and Microsoft worked to restore the status quo, the board ultimately reversed course, with the members opposed to Altman stepping down. The apparent failure of the non-profit board to influence the for-profit organization goes directly to Musk’s case that the transformation of OpenAI from research organization into one of the largest private companies in the world broke the implicit agreement of the organization’s founders. David Schizer, a former dean of Columbia Law School who is being paid by Musk’s team to act as an expert witness, echoed McCauley’s concerns. “OpenAI has emphasized that a key part of its mission is safety and they are going to prioritize safety over profits,” Schizer said. “Part of that is taking safety rules seriously, if something needs to be subject to safety review, it needs to happen. What matters is the process issue.” With AI already deeply embedded in for-profit companies, the issue goes far beyond a single lab. McCauley said the failures of internal governance at OpenAI should be a reason to embrace stronger government regulation of advanced AI — “[if] it all comes down to one CEO making those decisions, and we have the public good at stake, that’s very suboptimal.”
View
Perplexity’s Personal Computer is now available to everyone on Mac
Perplexity’sPersonal Computer, its answer to OpenClaw and other local AI agents, is now available to all Mac users via its desktop app, the companyannouncedon Thursday. As a reminder,Personal Computeris an expansion on Perplexity’s general-purpose, multi-model digital worker dubbed, confusingly,Perplexity Computer.“Personal Computer,” meanwhile, is designed to bring those capabilities to your own device. It does so by allowing AI agents access to local files, applications, and connectors, as well as the web, in order to handle the individual user’s personal, multi-step workflows. Or, as the company describes Personal Computer, it “takes Computer out of the cloud-only world and onto the device where most of your real work already takes place.” The goal is to capitalize on the growing demand for local AI agents, popularized by OpenClaw, which can perform tasks on users’ behalf. But while OpenClawpresented severalsecurity risksbecause of its elevated permissions, solutions like Personal Computer are meant to offer users a safer AI-enabled computing environment. (Or at least that’s the claim.) Perplexity’s Personal Computer wasintroducedlast month, but was limited to Perplexity Max subscribers and involved a waitlist. Today, the company says anyone on a Mac can now try the software as part of its new Perplexity Mac app. (Anyone can download the new app, but Personal Computer requires a Pro or Max subscription.) At launch, the software is able to work with your local files, native Mac apps, and operate on the web. It can also orchestrate tools, files, use over 400 connectors, and leverage your personal context, all within a secure development environment on Perplexity’s servers. If paired with Perplexity’s AI-powered Comet web browser, it can operate web-based tools without the need for direct connectors. Designed to run autonomous agents on an always-on device like a Mac Mini, Personal Computer can even be accessed remotely from your iPhone, where you can initiate tasks or approve requests from your device. Perplexity suggests this can be used for all types of work, like working with spreadsheets, documents, and projects with many different materials. Because the tool can work across apps, agents could do something like compare two files from different apps or pull notes from one app to create a draft in another. As a result of its general availability, Perplexity says its older Mac app will be deprecated in the weeks ahead so the team can focus on the Personal Computer app. The new Mac app is available as adirect downloadonly for now; it’s not in the Mac App Store.
View
OpenAI introduces new ‘Trusted Contact’ safeguard for cases of possible self-harm
On Thursday OpenAIannounceda new feature called Trusted Contact, designed to alert a trusted third party if mentions of self-harm are expressed within a conversation. The feature allows an adult ChatGPT user to designate another person as a trusted contact within their account, such as a friend or family member. In cases where a conversation may turn to self-harm, OpenAI will now encourage the user to reach out to that contact. It also sends an automated alert to the contact, encouraging them to check in with the user. OpenAI has faceda wave of lawsuitsfrom the families of people who have committed suicide after talking with its chatbot. In a number of cases, the families say ChatGPTencouragedtheir loved one to kill themselves—or evenhelped them plan it out. OpenAI currently uses a combination of automation and human review to handle potentially harmful incidents. Certain conversational triggers alert the company’s system to suicidal ideations, which then relay the information to a human safety team. The company claims that every time it receives this kind of notification, the incident is reviewed by a human. “We strive to review these safety notifications in under one hour,” the company says. If OpenAI’s internal team decides that the situation represents a serious safety risk, ChatGPT proceeds to send the trusted contact an alert—either by email, text message, or an in-app notification. The alert is designed to be brief and to encourage the contact to check in with the person in question. It does not include detailed information about what was being discussed, as a means of protecting the user’s privacy, the company says. The Trusted Contact feature follows the safeguards the companyintroduced last Septemberthat gave parents the power to have some oversight of their teens’ accounts, including receivingsafety notificationsdesigned to alert the parent if OpenAI’s system believes their child is facing a “serious safety risk.” For some time now, ChatGPT has also included automated alerts to seek professional health services, should a conversation trend toward the topic of self-harm. Crucially, Trust Contact is optional and, even if the protection is activated on a particular account, any user can have multiple ChatGPT accounts. OpenAI’s parental controls are also optional, presenting a similar limitation. “Trusted Contact is part of OpenAI’s broader effort to build AI systems thathelp people during difficult moments,” the company wrote in the announcement post. “We will continue to work with clinicians, researchers, and policymakers to improve how AI systems respond when people may be experiencing distress.”
View
Voi founders’ new AI startup Pit has become the latest rising star out of Stockholm
Swedish startupPitmay have gained notice for some rage-bait social media posts, but it has also become another Stockholm AI startup to watch. Pit is led by the cofounders of European scooter giant Voi including Voi CEO Fredrik Hjelm. He is joined by former iZettle and Klarna engineers. And it is now backed by a16z, which is leading the startup’s $16 million seed round. Stockholm,also home to Lovable, is one of the places where a16z has beenactively looking for the next European unicorn. Pit is going after enterprise AI with products intended to learn from the clients how their businesses run then create custom software to automate processes, Pit CEO Adam Jafer told TechCrunch. Jafer left Voi last summer after a seven-year tenure during which the company scaled into a team of nearly 1,000 employees operating into 13 countries. From his engineering viewpoint, Jafer saw how AI has matured enough for enterprise use. Initially, he saw a chance to replace low-hanging SaaS tools with in-house apps, but he soon envisioned an opportunity beyond Voi. “The aha moment for the bigger opportunity was when the models were no longer just chatbots that generate text, but became more agentic and could do things,” he told TechCrunch. Unlike competitors offering AI agent-building or vibe-coding products, Pit positions itself as an “AI product team as a service.” Pit is entering a crowded market and hopes to differentiate itself by relying on two pillars: Pit Studio, which lets enterprise employees guide it through processes that could be handled by AI-generated software; and Pit Cloud, which, the startup promises, provides that software in a way that meets enterprise requirements on governance, certifications and auditability. In mid-January, the startup started testing its plan with pilot customers in telecom, healthcare, logistics and other sectors, focusing solely on automating internal processes. “Nothing customer facing, no conversational AI, just pure back-office, service and support functions that we turn into automations so that you can give back time to people to focus on your core business,” Jafer said. The startup is now preparing to scale up commercially, but it won’t be hands-off. Following the trend of AI companies hiringforward-deployed engineers(FDEs) to embed themselves to drive enterprise adoption, Pit is also hiring solution engineers. The goal, Jafer said, is to meet the expectations of the large customers it is targeting. “They’re looking to buy outcomes. They want processes to go faster. They want to see productivity unlock and time unlock,” he said. Jafer said Pit is not pitching itself as a way to reduce human labor and cut jobs. “The theme is more around moving people upstream to do more valuable things for the business, rather than repetitive back-office work.” Success metrics also go beyond saving time and money. “Some of it is just quality of work improvement, reducing human errors and so on.” Yet Pit’s own needs on this became a subject of controversy a few months ago whenJafer posted on LinkedIndeclaring “Yes, our team currently has no junior engineers. At Pit, agents now do most of what junior engineers used to do.” While the post is still visible, he no longer stands by that. “It may have started like that, but you need a good mix as you scale,” he said with a smile. Hjelm anticipated the all-male team might raise eyebrows, too. In apost on X, he wrote that Pit was “founded by tech bros, from Voi and Klarna,” but immediately added, “We have tech girls on the team as well, fyi.” That clarification wasn’t immediately apparent from Pit’s LinkedIn profile, although TechCrunch has spoken with one woman working at Pit on the communications side. What the picture does reflect, though, is a sense of getting the band back together. Voi’s four cofounders haveremained friends over the years, and three of them are now part of this new journey: Hjelm, Jafer and Filip Lindvall, now a founding engineer at Pit. One of the startup’s engineers, Andreas Hjelm, is none other than Voi CEO Fredrik Hjelm’s brother. While Fredrik Hjelm is named as a co-founder of Pit, too, he is still Voi’s CEO, so his role will likely be less hands-on for the time being. Since going profitable in 2024, Voi has been considereda potential IPO candidate, andclosed 2025 with strong results. But his involvement as a well-connected entrepreneur could still open doors — and already has, with a16z. In atweet, Hjelm explained how a16z partners Alex Rampell and Gabriel Vasquez ended up leading Pit’s round. He became acquainted with Ben Horowitz, Gabriel Vasquez and Jen Kha “a few years ago when they came to Stockholm to understand what they could do for European tech. We stayed in touch. When it came to picking partners for Pit, we didn’t need the money to get going, but we wanted the strongest backers we could find. So we picked them, and they picked us.” Jafer also corroborated that Pit didn’t spend much time with other firms to raise its round, which was also backed by Pit’s founders themselves, as well as Lakestar, executives from American tech companies, and wealthy families from the Nordics. This transatlantic cap table confirms that there is growing interest for AI out of Stockholm, which has consolidated itself as one of the most active startup hubs in Europe.Pit could also benefit from its European DNA when it comes to sales. “We’re going after industrials, and there’s plenty of that in Europe,” Jafer said. He also reported that clients appreciate Pit’s agnostic approach. Since it can use different AI and cloud vendors depending on clients’ preferences, it could benefit from thecurrent tailwinds for sovereign tech, especially in critical sectors. “EU models running on EU compute is top of mind for almost every CIO we’re meeting,” Jafer said.
View
Google Upgrades AI Mode, AI Overviews With Expert Advice and Link Previews
Google has been frequently updating its artificial intelligence (AI) tools in Search, namely AI Overviews and AI Mode. After releasing agentic capabilities and new features in the Google app for Windows and Chrome browser, the Mountain View-based tech giant has now introduced five more features focused on usability and ease of access. Users can now check expert advice, easily view their preferred sources, get link previews, and see personalised deep dive suggestions. These new features are now rolling out to users globally.
View
Spotify wants to become the home for AI-generated personal audio
For the last few years, apps likeGoogle’s NotebookLM,Hero, and latelyAdobe Acrobathave given users the ability to create podcasts based on existing material like documents, daily schedules, and articles. Now, Spotify is letting you access these podcasts within its app, but you’ll need some programming tools to do so. The company said that if you already use a tool like OpenAI’s Codex, Anthropic’s Claude Code, or OpenClaw, you can useSpotify’s new CLI tool, which is in beta, to create the podcast and import it to Spotify to consume later. “People are already starting to use their agents to create personal audio that guides their day: from summaries of class notes before an exam to briefings of what’s on their calendar. And they’re asking for a way to listen to it on Spotify, where they already listen to everything else,” the company said in ablog post. The podcasts will appear in the user’s Spotify library for easy access, but aren’t available to other Spotify users. To take advantage of the new feature, users can go to the tool’s GitHub page and follow the instructions there. They would be prompted to log in to their Spotify accounts through a browser. Loading the player… After that, they can write a prompt like “Build me an audio session that dives deep into the history of the World Cup with details about key players, where it’s been held, and what I should know about the gamesthis year,” and ask the agent to generate a podcast and save it to Spotify. Users would also get a link to the Spotify listing for their podcasts.
View
China’s Moonshot AI raises $2B at $20B valuation as demand for open source AI skyrockets
Chinese AI companies may not be swimming in as much cash as their Western rivals, but their open source models are still facing no shortage of interest from those who don’t mind a performance hit in exchange for cheap inference. And investors are taking notice. Moonshot AI, the Beijing-based AI lab developing the popular Kimi series of open-weight large language models, has raised about $2 billion at a valuation of $20 billion, according to apostby Huafeng Capital, which advised some investors who participated in the round. The round was led by Chinese food delivery company Meituan’s VC arm, Long-Z Investment, a spokesperson told TechCrunch. Also participating were Tsinghua Capital, China Mobile, and CPE Yuanfeng, according to the post. The company raised $3.9 billion over the past six months, according to Huafeng Capital. Moonshot was valued at $4.3 billion at the end of 2025,per reports, and by early 2026, that figure had more than doubled to$10 billionfollowing a $700 million raise. Moonshot AI was founded in 2023 by Yang Zhilin, a former Meta AI and Google Brain researcher, and quickly became one of China’s most popular AI labs after itsopen-weight Kimi K2.5large language model took the coding world by storm earlier this year, nearly topping benchmarks and posting performance figures close to that of Open AI and Anthropic’s models at the time. The company’s latest model, Kimi K2.6, is currently the second-most used LLM on distribution platformOpenRouter. The fundraising comes as investor appetite for open-weight AI models made by Chinese labs surges. Moonshot’s annual recurring revenue topped $200 million in April, driven by rapid growth in paid subscriptions and API usage, per the financial advisor’s post. DeepSeek, perhaps the most popular Chinese AI lab, isreportedlyin talks to raise outside capital for the first time, at a valuation of about $45 billion. Some of Moonshot’s rivals have even gone public on the back of demand for their AI models. Zhipu AI,which trades in Hong Kong as Knowledge Atlas Technology, ended Thursday with a market cap of HK$434.7 billion (roughly $55.9 billion), whileMiniMaxended the day at HK$257.3 billion ($33 billion), after both stocks rallied on new model releases. Moonshot’s Kimi models compete with OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude, as well as ByteDance’s Doubao, Alibaba’s Qwen, Zhipu’s Z.ai, and DeepSeek. Moonshot’s backers include Alibaba, Tencent, HongShan (formerly known as Sequoia China), ZhenFund, IDG Capital, and 5Y Capital.
View
2 days left: Get 50% off a second pass to TechCrunch Disrupt 2026
Two days. That’s all that’s left to lock in your place — with your partner, co-founder, or colleague — atTechCrunch Disrupt 2026. Right now, you canbuy one pass and get 50% off a second of the same ticket type, but that offer ends May 8 at 11:59 p.m. PT. After that, prices go up, and the opportunity to show up with more perspective, more context, and more clarity disappears with it. At this stage, the advantage comes down to how quickly you leave with a clear sense of what to do next, which is whysecuring your pass nowand deciding who to bring with you matters more than waiting. Success in the startup ecosystem depends on knowing what to do next — and moving on it with confidence. Across founders, investors, and operators, the challenge isn’t a lack of ideas. It’s clarity. There are too many signals, too many opinions, and too many possible directions. Product decisions stall. Investment timing stretches. Execution slows, not because the path isn’t there, but because it isn’t obvious. Disruptcompresses that uncertainty into three days of high-impact programming, unparalleled networking, and real-time insight from the people actively shaping the market, giving you access to clarity that’s difficult to replicate elsewhere, and even harder to access if you wait past the May 8 deadline tosecure your second pass for 50% off. You’ll hear directly from leaders like: See who else is speaking in the growing lineup. One of the biggest advantages of being atDisruptis witnessing how decisions actually happen. Startup Battlefield 200makes that clear. As founders pitch live in front of seasoned VC judges and a global audience, you’re not just watching — you’re seeing what gets challenged, what resonates, and what ultimately stands out. That level of transparency is hard to replicate elsewhere, which is exactly why being in the room — and locking in your pass while you can stillbring someone with you for 50% off— matters more than trying to piece these signals together after the fact. What makesDisruptdifferent isn’t any one session — it’s how patterns emerge across them. You hear one perspective, test it in a roundtable, and see it reinforced — or challenged — in conversation later that day. Over time, the signal becomes clear. For founders, that might mean refining product direction. For investors, spotting what stands out. For operators, pressure-testing how to build and scale. Bringing a co-founder, operator, or partner accelerates that clarity. You compare interpretations in real time, challenge assumptions, and make better decisions while the context is still fresh — an advantage you can only lock in by securing your placebefore the 50% off a second pass offer ends. All passes are eligible for thebuy one, get one 50% off discount— so you can bring someone in your role or a complementary one and get more out of every conversation. But only if you act by May 8. Founder Pass— Made for startup builders. Access investor meetings, the Deal Flow Café, curated networking, and programming on scaling, fundraising, and growth. Investor Pass— Designed for VCs and angels. Connect directly with founders, access curated deal flow, and participate in investor-focused sessions and networking. Attendee Pass— Ideal for operators and builders. Full access to stages, breakouts, roundtables, and networking to understand what’s working across the ecosystem. Non-profit Pass— Tailored for mission-driven organizations. Explore how emerging tech applies to your work and connect with builders and partners. Expo+ Pass— Focused access to the Expo Hall, breakouts, and networking. Ideal for scouting talent, products, and emerging companies. Thesecond pass at 50% off dealends May 8 at 11:59 p.m. PT. If Disrupt is already on your radar, the decision now isn’t whether to attend — it’s whether you’re willing to move faster than the people who wait — especially when, for the next two days, you still have the opportunity to bring someone with you at 50% off.Register before this week ends to get these savings. Because once the offer ends, you’re not just paying more — you’re making your next set of decisions without the clarity everyone else is working from.
View
Aurora’s Chris Urmson on why self-driving trucks are finally ready to scale
Self-driving has been “almost here” for over a decade. But somewhere between DARPA challenges and a handful of driverless trucks hauling freight between Dallas and Houston, Aurora co-founder and CEOChris Urmson’s story changed. The self-driving truck company started commercial driverless operations last April and is now scaling from a handful of trucks to hundreds this year. On this episode of TechCrunch’sEquitypodcast, we’re bringing you a conversation Rebecca Bellan had with Urmson at theHumanXconference in San Francisco. The pair dug into the long road from lab to highway and how physical AI differs from the LLM boom everyone else is chasing. Listen to the full episode to hear about: Subscribe to Equity onYouTube,Apple Podcasts,Overcast,Spotifyand all the casts. You also can follow Equity onXandThreads, at @EquityPod.
View
Exhibit at TechCrunch Disrupt 2026: Get in front of 10,000 decision-makers before space runs out
You can spend months trying to get in front of the right people. Or you can show up where they’re already looking. AtTechCrunch Disrupt 2026, more than 10,000 founders, investors, and operators will gather at Moscone West from October 13–15. And at the center of it all — where deals start, conversations accelerate, and new companies get discovered — is the Expo Hall. If startup visibility, traction, and real conversations matter, you should be on the Disrupt exhibit floor, not on the sidelines.Get your exhibit table at Disrupt before your competitor does. Disruptisn’t a passive audience; it’s an active ecosystem. Investors, operators, and decision-makers don’t just walk the floor. They show up with intent. They’re there to: That intent is what makes the Exhibitor Program work. Instead of waiting to be discovered, you’re positioned directly in the path of: This isn’t just visibility. It’s proximity to opportunity. Exhibit tables are limited.Find out how to reserve an exhibit table. An exhibit table atDisruptis the entry point, but the value comes from everything around it. For $12,500, your team is set up for three full days in the Expo Hall, the highest-traffic area of Disrupt, with a fully branded 6’ table, signage, and seating included. More importantly, you’re equipped to operate beyond the booth. Your Disrupt exhibit package includes: That means your team isn’t limited to one spot; you’re moving through the event, joining conversations, and meeting people where decisions are actually made. You also get: Along with: Everything is set up to make sure you’re visible, credible, and easy to find.Reserve your exhibit table at Disrupt. Startups return to exhibit atDisruptyear after year because the results are tangible: qualified foot traffic, high-value conversations, and real opportunities. Startups come to: And they’re doing it inside a highly concentrated environment: That level of density changes the timeline. Conversations that might take months to initiate can start — and move — within days. Don’t be the last to reserve an exhibit table.See what you get as an exhibitor. The Expo Hall at Disrupt brings opportunity into clear view. This is where: Ideas move quickly here, from introduction to conversation to opportunity. If your company is building something real, this is where it gets seen.Reserve an exhibit table while you still can. The Exhibitor Program at Disrupt is open to startups at any stage and across any industry. It’s most valuable for teams that are ready to: If timing matters — and it usually does — this is a direct way to compress it.See exhibit details and deadlines. There is a limited number of 6’ exhibit tables available, and no extensions. They are: Once they’re gone, there’s no additional inventory. If you’re serious about growth, don’t wait. Disrupt is where startups get discovered, where conversations turn into capital, and where products find their market. The Expo Hall is where those conversations start. If you’re not on the floor, you’re not part of them.Join us as an exhibitor. You can spend months trying to get in front of the right people. Or you can spend three days at Disrupt and meet them all in one place. Secure your exhibit tablebefore space runs out.
View
Startup Battlefield 200 applications close May 27: A shot at VC access, global visibility, TechCrunch coverage, and $100K
Startup Battlefield 200applications are open, but only for three more weeks. Apply by May 27 for your shot at VC access, global visibility, TechCrunch coverage, $100,000 equity-free, and more opportunities for major scaling impact. Pre-Series A founders — and anyone who knows a startup worth backing — this is your reminder: The deadline is approaching fast, and the strongest contenders are already entering the arena. If your startup has been nominated, don’t wait.Complete your application now before the window closes. Know a startup that deserves to step into the spotlight?Nominate themnow to give them time to complete the application by the deadline. This is not just another pitch competition. Startup Battlefield 200 puts you on the main stage atTechCrunch Disrupt 2026in front of 10,000+ attendees, top-tier investors, media, and the global TechCrunch audience. You are competing live, getting direct VC feedback, and proving your company belongs among the next breakout startups. Waiting until the last minute is the fastest way to lose your edge. Early applicants have more time to prepare, more opportunities to stand out to the TechCrunch editorial team, and a stronger chance to make an impact before the competition intensifies.Nominate your startup, then finish the job and apply. We’re looking for ambitious early-stage startups building innovative, potentially category-defining products. Applications are open globally across every industry. Most selected companies are pre-Series A, though select Series A startups may qualify case by case. A functional MVP and clear product demo are required. Most importantly, we’re looking for founders building with vision, execution, and real market impact. This is the same launchpad wherecompanies like Dropbox, Discord, Fitbit, Trello, and Mint gained early momentum. Thousands apply every year. Only 200 are selected. Just 20 finalists pitch live on the Disrupt Stage. One startup takes the crown. Selected startups receive one of the highest ROI opportunities available to early-stage founders. It’s free to apply, and the potential return — from investor exposure to media coverage and customer growth — can create real scaling impact. Applications close May 27. The founders who break through are not waiting until the final hour — they are already making their move. If you are building something category-defining, or know a founder who is, now is the time to step forward.Nominateyour startup — or one that deserves the spotlight — andcomplete your applicationbefore the deadline runs out.
View
How Anthropic’s Mythos has rewritten Firefox’s approach to cybersecurity
When Anthropic unveiled its new Mythos model in April, it also delivered a stern warning to anyone developing software. The model was so powerful at sniffing out software vulnerabilities,the lab claimed, that it had discovered thousands of high-severity bugs that would need to be fixed before it could be made public. Now, security researchers for Mozilla’s Firefox browser are providing a closer look at what that process has looked like in practice, and what Mythos’ powers mean for software security at large. In a post published on Thursday, Mozilla said Mythos has unearthed a wealth of high-severity bugs, including some that had lain dormant in the code for more than a decade. That’s a significant improvement from what AI security tools were capable of even six months ago. Until now, AI bug-finding tools have come with severe drawbacks, often inundating security teams withlow quality reports and false positives. But Mozilla’s researchers say the latest generation of tools have turned a corner, particularly now that agentic systems can assess their own work and filter out bad results. “It is difficult to overstate how much this dynamic changed for us over a few short months,” the researchers wrote. “First, the models got a lot more capable. Second, we dramatically improved our techniques forharnessingthese models.” The results are striking: In April 2026, Firefox shipped 423 bug fixes, compared to just 31 exactly a year earlier. The researchers have also published details on 12 of the bugs, which range from a pair of unusual sandbox vulnerabilities, to a 15-year-old error in how the browser parses an HTML element. “These things are actually just suddenly very good,” Brian Grinstead, a distinguished engineer at Mozilla, told TechCrunch. “We see that on our own internal scanning, we see that on external bug reports, and we see that in all sorts of signals across the industry.” The fact that the system helped reveal vulnerabilities in Firefox’s “sandbox” system is particularly impressive, given how intricate an attack that exploits it needs to be. To find sandbox vulnerabilities, the model must write a compromised patch for the browser, then attack the most secure part of the software with the new code implemented. Finding and demonstrating the bug is a delicate, multi-step process, requiring both creativity and close attention. To put this into context,Mozilla’s bug bounty programpays researchers who can find a bug in Firefox’s sandbox up to $20,000 — the highest reward available. Despite the top-dollar bounty, however, Grinstead says Mythos is finding more sandbox issues than human researchers ever did. “We do get them,” he told TechCrunch, “but not at the volume that we are able to find with this technique.” Notably, the Firefox team still isn’t using AI to fix the bugs, despite well-documented progress in AI coding tools. The team does ask AI to code up patches for each bug, but the resulting code usually can’t be deployed directly, and instead serves as a model for a human engineer. “For the bugs we’re talking about in this post, every single one is one engineer writing a patch and one engineer reviewing it,” Grinstead says. “We have not found it to be automatable.” It’s still not clear how AI’s emerging capabilities will change the broader balance of power in cybersecurity. One month since Mythos was previewed, most of the bugs discovered likely haven’t been patched, which makes it hard to capture the full scope of their impact. Anthropic has been scrupulous about following responsible disclosure norms, but it’s likely bad actors are using similar techniques behind the scenes, even if the models they’re using aren’t quite as good. Speaking ata recent event, Anthropic CEO Dario Amodei was optimistic that the new tools would ultimately favor defenders. “If we handle this right, we could be in a better position than we started, because we fixed all these bugs. There are only so many bugs to find,” Amodei said. “So I think there’s a better world on the other side of this.” Having dealt with the gritty details, Grinstead has a more measured view: “It’s useful for both attackers and defenders, but having the tool available shifts the advantage a little bit to defense. Realistically, nobody knows the answer to this yet.”
View
