Four years ago, a startup lab launched that wasn’t quite an incubator, accelerator program, or venture firm. UP.Labs, as it was called then, built startups designed to solve problems for corporate customers such as Alaska Airlines and Porsche, as well as for the outside world.

The firm still has the same mission — albeit with a critical addition to its approach, a new name, and a $100 million investment from Silversmith Capital Partners. Vantora, as the firm is now called, continues to work with its corporate customers, including some new ones in industrial manufacturing that it declined to name, and in the oil and gas sector.

But now, it’s more focused on building startups solely for its corporate customers, and not for the broader market.

Founder and CEO John Kuolt told TechCrunch that Vantora is moving toward a “proprietary M&A pipeline.” This means Vantora will still build startups for its corporate partners, which invest in the ventures and serve as their first customers. But those corporate partners now have the option to fold the startups into their core businesses — and essentially keep them to themselves.

That shift has influenced Vantora’s increased focus on physical AI startups, according to Kuolt.

In the past, Vantora would end up spiking ideas that were strategic to its corporate partners, but too sensitive to bring to the outside world.

“We were missing on the biggest value problems, which had the biggest upside because of that,” Kuolt said in a recent interview. “Imagine you’re a Fortune 100 industrial company and you need to retrofit all of your hardware and machines for autonomy. You need to own that, it needs to be sovereign, and you can’t rely on a third party to go do that for you. You need to own that intelligence layer. They’re never going to let us go sell that to their competitors.”

This change has allowed Vantora to “unlock big physical AI use cases,” according to Kuolt, including with its existing customers.

For example, the firm came up with an idea to use AI to advance the business of its partner J.B. Hunt. “They said there is no way you can take this out to the world, and so we passed on it,” he said, adding that this proprietary model now allows Vantora to pursue it.

The firm launched in 2022 with Porsche as its first corporate partner. Since then, Vantora has launched several startups for Porsche and struck deals with Alaska Airlines, J.B. Hunt, Wabash, and TDG, the parent of Ashley Furniture.

In its early days, UP.Labs was tied — although never financially — to venture firm Up.Partners. While Vantora still shares office space with the California-based VC, it is its own entity, Kuolt explained, noting that the $100 million from Silversmith is the company’s first outside investment.

Anthropic has confirmed that it has a wet biology lab where it can use its AI models to run physical experiments.

AI leaders have been promising that AI is the key to curing human disease. Dario Amodei opined just last week: “I believe that AI could cure most major diseases in the next 5–10 years.” To do that, an LLM would have to have methods to test its theories in real life.

“We believe that to do biology, the final test is still, and will be for a while, in real lab work,” Anthropic’s head of life sciences, Eric Kauderer-Abrams told Reuters. “We absolutely are doing that today.” He added that the lab operates like most biotech labs: Anthropic conducts some research there while also working with externals partners.

This news probably shouldn’t be shocking. Anthropic bought Coefficient Bio, a stealth AI biotech, in April. It declined to say what the wet lab is working on, except that it’s not focused on drug discovery. Anthropic doesn’t want to give the appearance of competing with the pharma industry, where it has numerous major customers and partners. (For instance, it just announced a deal to work with Novo Nordisk on joint drug discovery.) Anthropic has already faced backlash for launching products perceived to compete with those of its customers.

To that end, Anthropic also launched a Life Sciences Verification Program this week, to give vetted bio researchers access to its most powerful models.

But the world is still reeling from the resignation of Anthropic researcher Jacob Coxon, who warned that “the people building AI earnestly believe that it could kill us all by the end of the decade.” Anthropic’s own alignment lead gave the odds at greater than 10% that AI could exterminate humanity within the next decade.

The fervor has grown so intense that CEO Dario Amodei published a post last weekend calling on the industry to slow down and institute self-regulation. He has repeatedly called bioterrorism one of AI’s biggest risks.

The juxtaposition of these dire warnings and the wet lab has not gone unnoticed in the tech industry. As investor and AI coding startup founder Chamath Palihapitiya posted on X, somewhat tongue-in-cheek: “The group behind such hits as: “We’re All Going To Die” and “Regulate Me Now” are building a wet lab in SF. I do not recommend this.”

Military aircraft were already in the air this spring when U.S. officials made an alarming discovery: the intelligence driving an armed operation against a Chinese vessel had been hallucinated by an AI chatbot. The operation was aborted at the last minute, narrowly averting a potential conflict with China, CNN reported on Friday.

The episode underscores a growing concern among military officials and outside experts: As decision-makers lean more heavily on AI, the errors these systems produce can travel up the chain of command before being questioned.

The intelligence report, which circulated during the war with Iran, said the vessel was carrying components for a nuclear weapons program.

The false intelligence originated with a Special Operations Command analyst who queried an AI chatbot to synthesize open-source data with classified signals intelligence. The chatbot misidentified the ship’s cargo manifest. The analyst then used the tool a second time to format the erroneous findings into an official-looking summary, which was circulated across command channels.

The near-miss comes as the U.S military races to integrate AI to accelerate decision-making and maintain its edge over China. The Pentagon has described AI as delivering a significant advantage in speeding up its kill chain so commanders can respond in the right time. But the same speed that makes AI attractive may also allow hallucinations with insufficient human oversight.

“It’s important for service members to understand the uncertainty inherent to LLMs,” said Jake Steckler, research scholar at GovAI and veteran U.S. Army officer in a written response to TechCrunch. “But it’s especially critical for any decisions that could lead to use of force, like targeting, intelligence analysis, or operational planning. There are life and death consequences for those decisions.”

Still, Steckler says, the incident should serve as a call to add more safeguards to AI, not a reason to avoid it. “These tools can be useful in the right contexts and with the right safeguards in place,” he said. “But prioritizing adoption speed over all else will likely lead to incidents that only make service members lose trust in these systems, which ultimately is only going to slow adoption.”

An aircraft equipped with Joby Aviation’s autonomy technology flew more than 3,100 miles across the United States without a human pilot taking control at any point, the company said Friday.

The flight not only demonstrated the capabilities of Joby’s autonomy tech, but also showed that the company’s ambitions extend well beyond its pursuit of electric air taxis. Joby said the autonomous aircraft is not done yet, and will make its way back to the West Coast with stops in various cities, including Louisville, Kentucky, and Salt Lake City, over the next two weeks.

Joby Aviation said the cross-country trip, which it described as the “first-ever autonomous flight across the United States,” included autonomous taxiing, takeoffs, navigation, and landings. The aircraft was remotely supervised from Joby’s headquarters in California and Shaw Air Force Base in South Carolina. A pilot was on board for compliance, but Joby said the aircraft performed the entire operation autonomously.

Joby, which used a converted Cessna Caravan aircraft for the trip, said the flight demonstrated that its autonomy technology could handle different environments and missions, including medical transport, disaster response, and defense logistics.

It also showcased the potential for a second business for Joby, which is best known for developing electric vertical takeoff and landing (eVTOL) aircraft for an eventual urban air taxi service. Joby has spent years developing and seeking type certification from the Federal Aviation Administration for its all-electric vertical take-off and landing aircraft, which will be used to carry passengers on short trips in urban areas.

While Joby’s electric air taxi service gets much of the attention, the company has a long-standing relationship with the U.S. Department of Defense and is increasingly pushing deeper into the military sector.

Its autonomy stack, which is designed to be retrofitted onto a wide range of conventional aircraft, is part of that strategy. The company plans to first bring to market a regional airframe equipped with the autonomy stack, which could be used for defense missions as well as commercial freight transport. That effort has been supported by a $17 million Air Force contract with AFWERX.

Joby took its most meaningful step into defense in August when it acquired Resonant Sciences for $500 million in cash and stock. Ohio-based Resonant Sciences, which makes radio frequency and sensor systems, is now part of a dedicated defense business within Joby that includes several projects such as the development of turbine-electric and hydrogen-electric aircraft, and an autonomy technology stack.

Last year, Joby signed an agreement with defense contractor L3Harris Technologies to “explore opportunities” to develop a new class of aircraft — specifically, a gas-turbine hybrid vertical take-off and landing aircraft that can fly autonomously — for defense applications.

Dario Amodei’s plans to put third-party safety evaluators inside AI labs are taking shape: Anthropic said that staff from technology consulting giant Accenture will begin working inside the company to scrutinize its models and staff.

In a blog post, Anthropic said that Faculty, a company Accenture acquired in January to act as its AI division, will begin “evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards.” Both companies expect to invest at least $1 billion in the project over the next five years.

The choice of Accenture surprised many AI watchers — and the markets, where the consultant company’s shares shot up 8% after hours. The discussion around embedded evaluators that sprang from Amodei’s blog post has focused on AI safety research organizations like METR, Redwood Research, and Apollo Research. That’s particularly true at Anthropic, which puts AI safety and alignment at the heart of its mission.

Anthropic said more evaluators will be announced in the weeks ahead and that it is in conversation with METR and other non-profit organizations about how to “pilot elements of embedded evaluation using their own funding.”

While Accenture is not known for its work on the bleeding edge of deep learning research, Anthropic pointed to the company’s practical experience deploying AI for large corporations and government agencies as key advantage. It is also, as a large public company that predates the AI revolution, more functionally independent of Anthropic and the let’s-say-complex ecosystem around the AI lab.

The lab noted that no standards yet exist for evaluators’ access or communications and that it expected its approach to evolve over time. While external evaluations are already a major part of the release of process for new large language models, recent incidents have raised the stakes: AI agents deployed by OpenAI and Anthropic have hacked into outside websites without raising alarms inside the labs.

Some critics calling for a more responsible approach to building artificial intelligence see Amodei’s scheme for self-policing the AI industry as a plan to evade accountability for the misbehavior of AI models. Anthropic insists that these evaluators “do not reduce our accountability, but help to make it more verifiable. The safety of our models remains our responsibility.”

In Brief

Automattic has a new interim chief financial officer: Jeremy Klaperman, the CFO of the company’s WordPress VIP Enterprise business unit. The news, shared internally on Friday, follows last week’s attempted ouster of Automattic CEO Matt Mullenweg, which ultimately resulted in the departure of the board members who voted him out and other executives, including then-CFO Mark Davies. Davies was briefly interim CEO before Mullenweg retook the position.

In a Slack post, Mullenweg shared that Klaperman had previously acted as Automattic CFO when Davies had been on sabbatical. He noted that the board — whose new members have yet to be announced — will still need to evaluate internal and external candidates for the position before a final decision is made.

In addition, Mullenweg said in the post that a candidate to become the company’s chief legal officer had also just verbally accepted the position, replacing Chief Legal Officer Andy Missan, who has also since left the company alongside Davies. TechCrunch recently reported that Davies and Missan had signed reciprocal severance deals during Mullenweg’s 33-hour leave of absence before he returned as CEO.

“I have 100% confidence in our cash and financial position,” Mullenweg’s post concluded, adding “There is still work to do, but everything is within our control and depends only on Automattic’s ability to execute.”

In subsequent updates, Mullenweg also said to “stay tuned” for board announcements, and noted that special advisor and creator of Wolfram|Alpha, Stephen Wolfram, will continue in his position. The company’s next board meeting is scheduled for September 23.

In Brief

These days, it’s rare to hear of a startup founded in 2019, or one not focused on AI agents, raising a hefty Series C. But health insurance startup Angle Health announced Friday that it raised a $200 million Series C, alongside a $400 million tender offer, at a $2.7 billion valuation. The round was led by Vitruvian Partners with participation from Town Hall Ventures, Blumberg Capital, Portage Ventures, PruVen Capital, and Y Combinator. The tender offer allows employees to cash out some of their shares, the company told TechCrunch. It expects the round to close later this month.

Angle Health, a winter 2020 YC alum, helps small businesses obtain and manage what’s known as “level-funded health plans.” These are plans that land between fully insured and self-funded plans. In fully insured plans, the carrier takes all the risks; such plans are more expensive but costs are predictable. With self-funded plans, the company covers the expenses and costs can be unpredictable. With level-funded plans, businesses make predictable payments to carriers, are insured against higher-than-expected costs, and can potentially save money by getting a share of the surplus back if expenses stay low.

Angle says these plans can make health insurance more affordable for small businesses. It’s an AI-powered platform that helps small businesses choose and manage plans, integrating with payroll and HR systems. The startup says it serves over 5,000 businesses and is profitable.

This week, I moderated a panel on world models at the All In conference (no relation to the podcast), and it gave me a chance to dig into one of the most mysterious corners of the AI world. The big players in the space are Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs — and while both have accumulated a lot of buzz and funding, they also rank pretty low on the trying-to-make-money scale.

At their core, world models are about automating spatial intelligence, so the field could head in lots of exciting and lucrative directions, from robotics to interactive video to more complex self-driving systems.

But when I started to press on where we would actually see the tech commercialized, things got foggy. The closest thing I found to an authority was Michael Rabbat, a co-founder of AMI Labs and the company’s VP of World Models, who joined me on the panel. But when I pressed him on exactly what the company was working on, he was cagey. “We’ll talk about it when we’re ready to talk about it.” Over email, he clarified, “We’re still in a research and building phase, so we’re not talking publicly about any product plans or timeline.”

To be fair, AMI is less than a year old, so it’s fair enough to keep quiet. But this sort of caginess extends to the whole world-modeling space. World Labs’ Marble is probably the most fully developed product in the space, and its demos range from straightforward media creation, building explorable environments for video games, or CGI effects. There are robotics use cases too, but the whole platform seems more designed to demonstrate capabilities.

That secrecy even extends to these companies’ suppliers. On the sidelines of the same conference, I spoke to Alex de Vigan, CEO of Physicl — a data supplier for the burgeoning world model business. He says he knows Physicl’s data has been useful for whatever they’re building, but he’s still in the dark about what exactly that is. “I wish they would tell us more. We could build more useful data if we knew what they were working on,” de Vigan told me.

Part of the mystery comes from how versatile world models are as an idea. The simplest version is a navigable map of the world, similar to the AI models that power self-driving cars. But the same modeling approach that helps a Waymo weave through traffic could also help a humanoid robot carry boxes, or turn a few minutes of video footage into an explorable environment. AMI has already dipped its toe in manufacturing, biomedicine, robotics, and even AI software for doctors through its Nabia partnership. Surely it won’t pursue all of those — but maybe one or two of them are standing out?

No one doubts that there are lots of viable businesses to be built on world model tech — and as long as it’s easy to fundraise, there’s no particular pressure to focus on one. In fact, there’s good reason not to. If AMI announced tomorrow that they had built a humanoid OpenClaw or a next-generation Hollywood rendering system, a lot of other labs would suddenly be very interested in the space. Soon, the lab would face potential competition from the other world model companies, the neolabs, and even OpenAI and Anthropic.

In some ways, it’s the flip side of all that easy fundraising. Your competitors can fundraise, too — and the same money that lets you build under the radar is also funding lots of potential rivals once the path to market becomes clear. But even if that competition is inevitable, it’s best if you delay it for as long as possible, which means keeping quiet about exactly what you’re building.

Cixin Liu fans will recognize this as a dark forest scenario: If you don’t know who else is in the woods, it’s best not to attract attention.

ChatGPT broke Diogo Almeida’s heart.

Almeida was an OpenAI researcher who helped build the chatbot and then invent reinforcement learning from human feedback (RLHF), the model-training technique perhaps most responsible for our current age of AI. But despite its capabilities, he was disappointed.

“We have lightning in a bottle, and yet it is not useful,” Almeida told TechCrunch. “I’ve been battling that problem since then. It took me a while to come to the conclusion: The problem is we are optimizing for human language … We have been super good at human language for four years, but it’s not useful for automation because computers speak a different language.”

Two years ago, Almeida left OpenAI to start TypeSafe AI, a startup trying to fix that problem. This week, the company released a new transformer-based model, Jev, that is not a large language model (LLM). It doesn’t output text, but instead produces probabilities, or what the company calls “calibrated decisions.”

Eschewing language does a few things: It makes the model incredibly cheap and fast, and because users define the outputs in advance, it cannot hallucinate. Its output tokens are free, and input tokens are metered by the billion, not the million.

A screenshot shows a comparison of Jev and an OpenAI model responding to the same requests Image Credits:TypeSafe AI

Developers are taking a great interest in the product; the company briefly lost the ability to serve users from its API because demand was so high. Jev appears most useful for software automation. Thus far, software developers see it as a cheaper and more robust way to incorporate intelligence into their code.

For example, Pranit Sharma, a software engineer at Vercel, a company making agentic infrastructure, said his company had used OpenAI’s ChatGPT Luna 5.6 to run a classifier to review commands for safety. When Vercel replaced OpenAI’s Luna with Jev, it got results five to 18 times more quickly and with greater accuracy.

Another developer, Bryo AI CTO Nikhil Mudholkar, tested Jev against Gemini for classifying business emails. In his test, Gemini was slightly more accurate, but 10 to 20 times more expensive. More interesting to Mudholkar were Jev’s confidence scores — “it is the only one that hands back a real probability which makes it ideal for automating workflows!!”

Besides replacing LLMs in certain use cases, the new model can also augment them, acting as a smart check on misbehavior. Using agents to monitor agents can quickly become expensive, but using Jev to do so, Almeida argues, makes sense. He sees users deploying Jev to track LLM agent traces and prevent jailbreaks.

“At the end of the day, it delegates the hallucination problem a little bit to the user,” explained Armin Ronacher, the CTO of Earendil, which builds the open source model harness Pi. “The user has to say, okay, if this only comes back with 50% probability, maybe this is a coin toss, and I disregard it. But if it’s 95%, sure, then I can do something with it.”

Another potential use for Jev is model routing, Ronacher said. Predicting whether a given workload requires a specific model would be useful, but using an LLM for the job would be expensive. Jev’s low cost and speed make that kind of real-time sorting possible.

And that’s Almeida’s hope. The model is named after William Stanley Jevons, the 19th-century economist whose eponymous paradox describes how the falling cost of a commodity can lead to it being used more and more. In this case, the falling cost of intelligence should lead to its widespread deployment.

“We think that there’s just going to be smart software all over the place in a way that’s emergent and distributed … much more like the early internet than you know like the mega apps that people are trying to build right now,” Almeida said.

Almeida is tight-lipped about the model’s architecture, which outside observers suspect is built on top of an open-weight LLM. The company refers to Jev as a “System One model,” focused on intuition rather than reasoning, and specifically focused on the right task. Almeida says Jev is trained exclusively on synthetic data using a technique he calls “reinforcement learning from calibrated decisions.”

“We made an early bet that we will be making all of our data, and that has been one of the best bets I’ve ever made in my life — better than our launch, in my opinion, better than RLHF,” he told TechCrunch. “Half of [our company] is a lab that basically owns this entire subfield of statistically well-understood synthetic data, and that is now my life joy.”

For now, Jev stands alone as this kind of model, but Ronacher expects that competitors will spring up now that its utility is apparent.

“We should have seen this earlier in many ways, but presumably because the LLMs are so cheap and subsidized, you often don’t have to be creative yet,” he said.

TypeSafe itself will be building more versions of the model, in new modalities. Asked if TypeSafe is a frontier lab, Almeida said, “the main product of frontier labs is fear or hype. I would like our main product to be intelligence…[but we are] not a lab in the sense of, you know, like bet on infinite wealth, or a religion, or building God in a data center, or whatever is the thing of today.”

In Brief

Disney has hired its first-ever chief technology officer and in a curious twist, the new executive hails from an AI startup that the Magic Kingdom previously accused of infringing on its IP.

Karandeep Anand is the former CEO of Character.AI, a company that has weathered numerous legal problems since it was founded in 2021. One of those legal problems arose in September 2025, when Disney sent the company a cease and desist letter accusing it of infringing on its beloved characters.

Character.AI allows users to create distinct virtual characters with generative AI and talk and interact with them. Disney previously claimed that the company was hosting copyrighted characters from its franchises.

In addition to Disney’s accusations, Character.AI has also been sued over allegations that the company’s chatbots encouraged users to commit self-harm and suicide.

Variety reports that Anand was chosen for the role by new Disney CEO Josh D’Amaro, who took over after former company chief Bob Iger stepped down in March. The hiring suggests D’Amaro wants the company to embrace new technologies.

Anand formerly served as a board adviser to Character.AI before becoming CEO in May 2025, according to his LinkedIn profile. He also worked at Facebook between 2015 and 2021 and, before that, spent 15 years at Microsoft.

For some iPhone buyers in India, getting Apple’s brand-new iPhone 18 Pro or iPhone 18 Pro Max did not require lining up outside a store or waiting days for delivery. They needed only a few taps on a quick-commerce app to get one delivered to their doorstep in about 10 minutes.

Blinkit, Zepto, Swiggy’s Instamart, and BigBasket — platforms best known in India for delivering groceries and everyday essentials in minutes — began offering the new Pro models as they went on sale in the country on Friday. The phones are, however, also available through Apple’s usual sales channels, including its stores, authorized retailers, and online marketplaces.

Putting a new iPhone on a 10-minute delivery app is not entirely new in India. Quick-commerce platforms have offered Apple’s latest devices on launch day previously as well, as they have steadily expanded beyond groceries into categories including electronics, beauty products, and fashion.

However, the market behind those deliveries is growing rapidly. India’s quick-commerce sector reached about $9 billion in the first half of 2026, per a recent report by Redseer Strategy Consultants, with the number of monthly users rising to more than 60 million from 8 million three years ago. Google and Deloitte separately estimate the market could reach $50 billion by 2030, with non-food products such as electronics, fashion, and beauty accounting for 45% of spending.

There were signs of strong demand on launch day. BigBasket said it delivered more than 300 iPhone 18 Pro and Pro Max units in the first hour of sales. Instamart told TechCrunch that Bengaluru led the iPhone 18 Pro-series orders on its platform, followed by Delhi. The Burgundy 256 GB iPhone 18 Pro was Instamart’s most popular variant, accounting for nearly a third of its sales.

Nonetheless, within hours of launch, availability of the new iPhones became patchy. TechCrunch found the iPhone 18 Pro models sold out on some quick-commerce platforms and in some locations, while they remained available for delivery elsewhere in cities including Delhi and Bengaluru.

Tarun Pathak, research director at Counterpoint Research, said the launch-day deliveries showed that quick commerce could work as a channel for premium smartphones in India’s biggest cities, even though volumes remain small compared with overall iPhone sales.

The ability to fulfill high-value smartphone orders provides an “early proof of concept,” Pathak told TechCrunch. “The key question is whether this launch-driven demand can translate into regular smartphone purchases through quick-commerce platforms.”

Navkendar Singh, associate vice president at IDC, said India stood out globally for this kind of quick-commerce delivery and was particularly suited to such purchases because consumers in major cities had built trust in the platforms by using them regularly for everyday purchases. Official tie-ups between brands and the platforms also gave consumers greater confidence in buying expensive products through them, he told TechCrunch.

For a product like the iPhone, the appeal also comes down to “instant gratification,” Singh said, allowing buyers to avoid queues outside stores or waiting several days for an online order to arrive. Still, he cautioned that quick commerce was not yet a significant sales channel for smartphones. Its role around new iPhone releases is largely a “launch-day marketing phenomenon” that “creates good noise around the launch,” he said.

For the wealthy families who manage their own money through a family office, the math right now is simple, according to Djoann Fal, a family office advisor and investor at the private wealth platform Atlas Capital in San Francisco.

A family office might want to put money into something like green energy for the long haul. But in the current environment, if a fund manager offers a deal that could triple an investor’s money over three years, and another deal could triple it in three months, the choice is easy. As Fal put it: If they have one deal that has the chance to make them 3x in three years, and another deal that could make them 3x in one quarter, “they’re just going to invest in the AI deal that does 3x in three months.”

It isn’t surprising that family offices want in on the action, given that everything around AI is so hot right now, from valuations to pricing to the potential for returns. A bigger shift is that they no longer necessarily want to invest the traditional way, through a venture capital fund manager. Instead, family offices are increasingly buying existing shares in a private company from existing shareholders, or making direct deals on their own. Both approaches get them exposure to the hottest companies without having to hand over control of their money to a fund manager for a decade.

“They have more dry powder to chase single-name deals,” Fal said, referring to family offices increasingly skipping so-called blind-pool fund commitments, where investors hand money to a fund manager without knowing in advance which companies it will be used to back. He said a new generation of family offices is emerging, too, with a higher appetite for risk than their predecessors. “Right now, the single names they want are the AI leaders,” he said.

They definitely have the money. Family offices were overseeing $5.5 trillion in wealth as of 2024, according to a Deloitte report published that same year, which projected that number to hit at least $9.5 trillion by 2030.

Underscoring Fal’s point, they want to pour more of it into riskier, but potentially more lucrative, assets, too. UBS’s 2026 Global Family Office Report, which surveyed 307 family offices worldwide with an average net worth of $2.7 billion, found that alternative investments — including private equity, venture capital, and private credit — now make up 42% of the average family office portfolio.

Whether that’s a permanent shift or part of a familiar cycle is worth asking, though. Family offices have been here before, more than once. Direct investment activity climbed steadily through the late 2010s, for example, then spiked hard in 2021, when direct deals hit 13% of the average family office portfolio, up from 9% in 2019, according to UBS’s own tracking at the time.

Total family office deal activity peaked that year too, according to PwC’s Global Family Office Deals Study, at 17,460 deals worth roughly $1.05 trillion globally. Then it reversed just as fast, as rising interest rates and disappointing returns on some of those direct bets led many family offices to pull back. (Direct and M&A deal activity fell by 53% in just 18 months by late 2023, per PwC.) By the first half of 2025, overall family office deal volume had fallen to its lowest point in a decade.

Now it’s bouncing back, just as some would argue the market is overheated, and this time, family offices are writing bigger checks on fewer deals. Much of that activity is happening in the secondary market, which Fal calls the most “de-risked” asset in venture capital right now, since investors are often backing companies that already have customer traction and revenue proof. Angelina Hu, head of investor relations at Bridge Funding Global, put it another way: The secondary market lets family offices gain exposure to a private company without “taking exposure across 20-30 companies.”

“Family offices see the risk, but don’t want to miss the opportunity,” added Bruce K. Lee, founder of Keebeck Wealth Management.

That appetite is translating directly into Fal’s own business. He said he’s never raised this much money in his life — capital all earmarked for “AI things,” while clients who aren’t pouring money into AI are struggling to raise money at all. He said family offices currently have no problem paying “primary-style prices” for “secondary-stage risk,” even if it leaves them with “little room for outsized returns.”

Illustrating his point, Fal, whose clients primarily back climate-focused investments, said he spent the summer fielding interest from people looking to invest $50 million to $100 million into Anthropic through the secondary market. Anthropic stakes, alongside those of OpenAI, are among the most sought-after in the industry right now — what Emily Zheng, a senior VC analyst at PitchBook, described to TechCrunch as some of the “most contested real estate in venture.” Said Fal, “Even advisors with non-AI mandates are being pulled into transactions. That is where LP demand is.”

They see it as a comparatively safe investment in a volatile market. Indeed, a February report from J.P. Morgan Private Bank found that 65% of global family offices plan to “prioritize AI investments” despite concerns over inflated valuations and pricing. “Family offices are operating against a backdrop of geopolitical tensions, rising global debt levels, recession risk, and broader market uncertainty,” Maximilian Kunkel, chief investment officer of global family and institutional wealth at UBS Global Wealth Management, told TechCrunch.

Family offices clearly believe there is plenty of upside, too, in these investments. “AI is viewed as one of the most powerful long-term growth opportunities,” he continued. “The result is that investors are not choosing between resilience and growth. Instead, they are maintaining exposure to AI while diversifying across regions, currencies, and asset classes to manage concentration and macroeconomic risks.”

Whether that’s diverse enough is the big unanswered question. If the AI boom turns out to be a bubble that eventually bursts, the fallout won’t stay contained to AI stocks. As Lee put it, “everybody knows that if it goes under, the stock market’s going to have issues.” He said there are two types of investors right now: those sitting on the sideline, calling AI a bubble without owning any stock, and those who are anxious but effectively all-in. Some of his clients, he said, talk about AI being a bubble but don’t like to hear about ways to hedge against it. “We’re all addicted to returns,” he said. “That’s the sugar.”

Cybersecurity teams with the U.S. Coast Guard and the FBI boarded two U.S.-bound oil tankers last month after hackers reportedly compromised at least one of the ship’s networks and took control of its navigation, propulsion, and cargo systems.

According to a joint statement shared with TechCrunch, the agencies boarded the vessels in the Gulf of Mexico between August 21 and August 24 to “ensure integrity of the vessel’s operational and information technology systems following indications that the networks of both vessels were compromised.”

The Coast Guard released several photos of its staffers and FBI agents climbing aboard one of the vessels, which were shared with CBS News. The statement said that the captain, crew, and the on-shore staff of the vessel’s owner cooperated with the agencies.

The incident is an example of how shipping vessels, oil tankers, and the broader global logistics network can be targeted by hackers. Tankers often have complex networks sprawling the span of the vessel to aid with navigation, organizing cargo, and ensuring the regular running of services necessary with operating a massive ship.

A spokesperson for the Coast Guard did not comment beyond its emailed statement, including naming the vessels involved.

CBS News confirmed that one of the tankers is VL Prosperity, a 333-meter oil tanker that can hold over 2 million barrels of oil. According to VesselFinder, the tanker is currently located in the Gulf of Mexico.

Citing Iranian media, CBS said hackers compromised the ship on August 7 while traveling from Egypt to the United States and interfered with the ship’s speed and fuel systems. The tanker also lost communications for more than a day, per CBS.

It’s unclear who is behind the attacks, but CBS News said the U.S. is looking into whether Iran is behind the hack. Iran has been behind a multitude of hacks in recent months following the start of the U.S. and Israel-led war against Tehran, which killed Iran’s supreme leader in February. Iranian-backed hackers have since launched a destructive hack at medical device maker Stryker, hacked the mass transit system in Los Angeles, and compromised over a hundred water facilities across the United States in response to the widespread U.S. bombing of Iran. CISA has called the attacks “opportunistic” in nature.

The joint FBI-Coast Guard statement said that there were “no reports of operational disruptions, vessel instability, physical danger to crews, or environmental impacts.”

You have an AI product to build. Do you choose a proprietary frontier model and get moving quickly? Build on an open model and gain greater control? Fine-tune your own version? Run locally? Use multiple models? Change strategy six months from now when the economics and capabilities shift again?

There may not be one right answer. But for founders, choosing badly can affect almost everything that follows — cost, infrastructure, margins, differentiation, speed, and control.

That decision is at the center of the session “The Open vs. Closed AI Debate Is Just Getting Started,” coming to the Builders Stage at TechCrunch Disrupt 2026, happening on October 13-15 in San Francisco. Led by Nvidia‘s Nader Khalil, Director of Developer Tech, and Sydney Sykes, Global Head of VC Partnerships, the session will discuss the trade-offs between open and proprietary AI and whether either approach can provide a lasting competitive advantage.

Image Credits:TechCrunch

This isn’t a philosophical argument about open source. It’s a business decision being made right now inside startups of every size. Dive deep into this AI debate with 10,000+ tech leaders at Disrupt by getting your pass. Register now and save up to $200 before prices go up on September 25 at 11:59 p.m. PT.

The AI gap is closing — the decision isn’t getting easier

Open models have advanced quickly. Nvidia said in July that 145 papers accepted at ICML 2026 cited its Nemotron open models and datasets, alongside research using other Nvidia open model families across robotics, autonomous vehicles, and biomedical research.

At the same time, proprietary frontier labs continue pushing model capabilities forward. The result is a market where the question is increasingly less about whether open models can be useful and more about where each approach makes commercial sense.

Even Nvidia rejects a simple either-or framing. At GTC earlier this year, CEO Jensen Huang argued that the future is not proprietary versus open, but proprietary and open. That sounds straightforward until you have to build a company around the decision.

If two models can deliver similar results, does lower cost win? What if one gives you more control over your data? Does owning more of the stack create defensibility, or just infrastructure you now need to maintain? And if the best model changes every few months, how tightly should your product be tied to any one of them? These are the questions Khalil and Sykes will unpack at TechCrunch Disrupt 2026. Don’t miss it — register for your ticket now to get $200 savings before September 25 at 11:59 p.m. PT.

Two perspectives shared on the Builders Stage on the same AI stack

Nader Khalil approaches the discussion from the builder and infrastructure perspective. Before becoming Nvidia’s Director of Developer Tech, where he leads open source and local AI, he co-founded Brev.dev, an AI infrastructure company acquired by Nvidia in July 2024.

Brev.dev was built around simplifying access to GPU infrastructure across different environments. Nvidia’s developer documentation described its tools as allowing developers to deploy AI software across public cloud, private cloud, and on-premises infrastructure without locking themselves into a single compute source.

Sydney Sykes, meanwhile, brings the venture ecosystem into the discussion as Nvidia’s Global Head of VC Partnerships.

Together, that creates room to examine the same decision from different directions: what developers need to build and what companies need to become investable, scalable businesses.

Sit front and center at one of the biggest debates in today’s world of AI. Register for your ticket before the savings of up to $200 end on September 25 at 11:59 p.m. PT.

Your model isn’t your moat — until it is

There is another uncomfortable question behind the open-versus-closed debate: Where does your competitive advantage actually live?

If competitors can access the same proprietary API, differentiation needs to come from somewhere else — proprietary data, workflow, distribution, customer relationships, product experience, or specialized technology. But choosing an open model doesn’t automatically give you a moat either.

You gain flexibility and potentially greater control, but you also take on decisions around deployment, optimization, and infrastructure. And the economics can change depending on the workload and scale.

Nvidia is investing heavily in that open ecosystem. Its Nemotron 3 Super, launched in March, is an open 120-billion-parameter model designed for agentic workloads, and companies are already combining it with proprietary models rather than treating the two approaches as mutually exclusive. That hybrid reality may ultimately be the most interesting part of the debate.

Register now to join the conversation on the Builders Stage in October. Savings of up to $200 end on September 25 at 11:59 p.m. PT.

What should you build on? Decide at TechCrunch Disrupt 2026

That’s what makes this session useful beyond an audience of AI engineers.

If you’re a founder, the choice can shape your margins, fundraising story, and product roadmap. If you’re an investor, understanding where value sits in the stack can help distinguish genuine defensibility from a thin product layer sitting on somebody else’s model. If you’re a line-of-business lead, it affects procurement, security, data control, infrastructure, and the freedom to change providers later. And for developers and students, it’s a chance to understand where the technical decisions being made today connect directly with the business models being built around them.

No one needs another abstract argument over whether open or proprietary AI is philosophically better. What builders need is a clearer understanding of the trade-offs.

Join Nader Khalil and Sydney Sykes on the Builders Stage at TechCrunch Disrupt 2026 and decide which side belongs in your AI strategy. Register now to save up to $200 before prices go up on September 25 at 11:59 p.m. PT.

Meta’s new AI assistant app, Muse, is now available on Mac.

On the Mac, Muse can interact with your files, messages, calendar, notes, and mail, all within their native applications.

The new app follows the launch of Muse on mobile and the web earlier this month, when it quickly rose to the top of the U.S. App Store charts.

muse for mac is HERE!! your agent can now get stuff done right on your computer — files, messages, calendar, notes, all of it.



you're in control of what it can access, and it always asks before doing anything sensitive.



Add it to the dock – https://t.co/bFHgBZlkLj pic.twitter.com/mecGVDzXiq — Alexandr Wang (@alexandr_wang) September 17, 2026

Just like on other platforms, you can control what Muse has access to on your Mac on an opt-in basis. In addition, the app will always ask for your approval before it performs sensitive actions, Meta says.

The desktop app comes at a time of intensifying competition between consumer-facing AI agents, which many in the industry believe will be the future of computing. The movement is disrupting the app industry, software-as-a-service businesses, and perhaps, eventually, even hardware, like the iPhone.

In addition to Muse, all the major players offer consumer-facing agentic AI. And other AI agents have found traction by letting people use AI agents via text messages. One of the first breakout hits in this space, Poke, exited to Cognition in July. Meanwhile, the newest AI assistant frequently brought up in tech circles these days, Instinct, is said to be raising more funds at a $10 billion valuation.

For now, the question is who can capture the most consumer market share in this space, and quickly. That’s led Muse and Instinct to rush to launch new features at a rapid pace — both rolled out voice calling this week, for instance.

As Mark Zuckerberg noted on X when announcing the Mac app, “The team is shipping fast.”

When Orion asked if I wanted to test its $2,195 AI-enabled mattress pad, I agreed purely because I thought it would be funny to write about. I mean, come on. What could possibly be “AI-enabled” about something you sleep on? And why does it cost more than my monthly rent in a major city?

The joke, as it turned out, was on me. Over the few months that I’ve had the Orion sleep system, I’ve been sleeping like a baby.

Orion is basically a cooling mattress pad that you can control using your phone, and it adjusts temperature as you sleep to minimize how often you wake up in the middle of the night. Sleeping on the pad feels like flipping your pillow to find “the cool side,” except that now your whole bed is the cool side of the pillow.

As for the AI aspect, Orion supposedly learns from your sleep data to modulate temperatures and maximize the quality of your sleep, though I’m skeptical of how necessary these AI features actually are (and how much they contribute to the hefty price tag).

By default, the Orion will start cooling down around the time you go to bed, which you manually input. It will then get colder as you fall asleep, and then when it’s time for you to wake up, it slowly brings the mattress pad back to room temperature, making it easier for you to get out of bed.

Image Credits:Orion Sleep

Orion’s founder Harry Gestetner is only 26 years old, but he’s more experienced as an entrepreneur than his age might suggest: He sold his last company Fanfix, a creator monetization platform, for an estimated $65 million when he was 21 years old. It’s an impressive success story, but these tales are never so glamorous behind the scenes. The struggles he faced while building Fanfix inspired this new venture.

“I was in the office seven days a week, totally neglecting every other area of my life, neglecting my health, sleeping terribly, eating terribly, not working out, and I started to get health problems,” Gestetner told TechCrunch. “I got this inner ear condition, which basically incapacitated me from working for a period of time. As someone who basically had no hobbies and likes to work seven days a week, that was very difficult.”

Gestetner got interested in the longevity movement, which is growing in popularity among some startup or productivity-maxxing types. While he is no Bryan Johnson, it does sound like Gestetner is sipping the Kool-Aid a bit. He tells me that he thinks we’re five years away from “longevity escape velocity,” which means that “for every additional year you’re alive, AI science is able to add an additional year to your life.”

It’s a tough sell, but that’s not what Orion is about anyway. The product is trying to convince you of something that you probably already know: You generally feel better after a good night’s sleep. Even Johnson, who has gone to such extremes as infusing himself with his son’s plasma, says that “Sleep is the world’s most powerful drug.”

“I think most people just want to feel good, and you know, be able to go play soccer with their grandkids up into their older years,” Gestetner said. “I think sleep is the cornerstone of longevity.”

Image Credits:Harry Gestetner, founder of Orion Sleep

Despite arriving in two intimidatingly large boxes, Orion was easy to set up. It looks like any sort of plush mattress pad, except it has tubes that connect to a tower, which resembles a chic bedside table from West Elm.

Once that’s done, you download the Orion phone app and work through its setup flow, which involves pouring water into the control tower, which is how the cooling system works. I haven’t had to re-up the water supply yet after months of use, which is impressive. If you share a bed with someone, you can even split the Orion into two “zones” that follow their own unique cooling schedules.

From there, you can simply forget that the Orion is there, except for when you get into your perfectly cool bed each night. It requires very infrequent maintenance.

And if you want, you can use the app to check data about your sleep, like how long you spend in light sleep, deep sleep, or REM, and how often you wake up or toss and turn in the middle of the night.

Speaking of data, it’s worth noting that such intimate health information (i.e., exactly when and how you sleep, in detail) is usually transmitted over the cloud for processing, and so may be compromised if a third party that’s involved in the connection gets hacked. Gestetner says Orion invests heavily in security, but this is still an inherent risk in using a connected health product.

Image Credits:Orion Sleep

While I can confidently say that the Orion has helped me sleep better, there’s no way I’d ever spend $2,195 on something like this. But I would probably feel differently if I were a Silicon Valley engineer making $500,000 a year.

However, in terms of limiting my phone usage before bed, the $59 Brick has done as much to make me sleep better as the Orion, though the products obviously do very different things.

Gestetner may be a millionaire after his last exit, but he still understands Orion’s price puts it out of reach of most people (it’s less expensive than competitor Eight Sleep’s $3,000 model, at least). He says he is working on a more affordable model.

“Eventually we want to get to a $500 price point product,” he said. “To use the Tesla analogy, this product is our Roadster, and we want to eventually get to our Model 3.”

The app you use to trade stocks increasingly wants to be the app you use to manage your financial life.

Investing, banking, credit, crypto, retirement, prediction markets, and AI-powered trading are converging. At the same time, consumers expect financial products to be as intuitive and responsive as the best technology products they use every day — without compromising the trust they expect when their money is involved.

Few companies illustrate that shift as clearly as Robinhood. At TechCrunch Disrupt 2026, Abhishek Fatehpuria, Vice President of Product Management for Brokerage at Robinhood, joins the Smart Money Stage with “Winning the Modern Financial Consumer.” He’ll explore how technology and changing expectations are reshaping financial services, and what it takes to build trusted products at massive scale.

Image Credits:TechCrunch

Join the conversation with Abhishek Fatehpuria on October 13–15 at Moscone West in San Francisco, where 10,000+ founders, investors, and tech decision-makers will come together to examine and connect with the companies, technologies, and business models shaping what comes next.

Register now to save up to $200 before prices go up on September 25 at 11:59 p.m. PT. Save up to 30% on group passes of four or more.

From trading app to financial platform

Robinhood, which became synonymous with commission-free stock trading, now spans a much wider financial landscape. Its offerings include investing, banking, credit, crypto, retirement, and prediction markets, while its recent launches have pushed further into AI, private markets, and international expansion.

The scale is significant. Robinhood reported 28.6 million funded customers and $384 billion in total platform assets at the end of August 2026.

And customers are increasingly using products outside Robinhood’s original trading proposition. In its second-quarter results, the company said Robinhood Banking had attracted more than $3 billion in deposits, its credit card business had surpassed $100 million in annualized revenue, and prediction markets had traded more than 3.5 billion contracts through that point.

The bigger story isn’t simply the number of products. It’s the attempt to earn a larger role in how customers manage their financial lives. That creates a very different product challenge. Learn all about navigating this evolution in fintech on the Smart Money Stage in October by registering for your pass today. Save up to $200 before September 25 at 11:59 p.m. PT.

Image Credits:Cindy Ord/Getty Images for Robinhood

The next customer may expect something completely different

Consumer expectations rarely change one feature at a time. Once people become accustomed to instant access, personalized recommendations, and seamless digital experiences in one part of their lives, they begin expecting them elsewhere.

Finance is no exception. Robinhood is now experimenting with AI-powered investing tools, including Cortex and Agentic Trading. In Q2, the company said nearly 100,000 customers had already opened Agentic Trading accounts, allowing users to trade equities, options, and crypto through AI-powered agents.

Meanwhile, prediction markets have become one of its fastest-growing businesses. Robinhood said 4.7 billion event contracts were traded in August alone — up 15x year-over-year.

The question for the wider market is what these behaviors tell us about the next generation of consumers — and which expectations will spread beyond Robinhood. Dive deep into what’s ahead in the finance world with one of the leaders in fintech by registering for your Disrupt pass. Register today to save up to $200 before prices increase on September 25 at 11:59 p.m. PT.

Whether you build, fund, or buy technology, you’ll find answers at TechCrunch Disrupt 2026

For founders, Fatehpuria’s session offers a look at how a product organization keeps expanding without losing sight of the customer experience that drove its original growth.

For investors, it’s a window into how one company is trying to increase wallet share by moving from individual products toward a broader financial ecosystem.

For line-of-business and product leaders, the lessons extend well beyond fintech: prioritization, trust, customer behavior, and the difficulty of adding complexity without making the experience feel complex.

And if you’re earlier in your career, studying where technology and finance are heading, it’s a chance to see how one of the sector’s best-known platforms thinks about building for the next generation of customers.

The financial consumer is changing. The products competing for that consumer have to change with them.

Join Abhishek Fatehpuria at TechCrunch Disrupt 2026 and hear what Robinhood is learning along the way. Register now to save up to $200 on your pass before September 25 at 11:59 p.m. PT. Save up to 30% on group passes for four or more.

Jeff Lawson has already helped build a company with more than $4 billion in revenue and 300,000 customers worldwide. After scaling Twilio into a public company, his next challenge reaches well beyond software: bringing fusion power to the grid as co-founder and CEO of Inertia.

It’s a striking change in direction. It also raises a question worth exploring: How much of what makes someone a successful software founder carries over when the next business depends on lasers, manufacturing, and major engineering milestones?

At TechCrunch Disrupt 2026, Lawson will explore that transition in “How Twilio’s Founder Is Tackling the Power Problem.” The fireside chat joins the Smart Systems Stage, where you’ll find the tech infrastructure programming, which includes climate and energy tracks, examining what changes when an experienced entrepreneur takes on a capital-intensive deep tech company.

Image Credits:TechCrunch

Join us October 13–15 at Moscone West in San Francisco, alongside 10,000+ founders, investors, and tech decision-makers. Get a closer look at the company-building decisions behind an ambitious energy venture. Register now and save up to $200 on your pass before rates increase on September 25 at 11:59 p.m. PT. Save up to 30% when you register as a group of four or more.

A $450 million raise is only the beginning

In February, Inertia announced a $450 million Series A, led by Bessemer Venture Partners, with participation from GV and other investors. The funding supports the development of powerful laser systems and the mass manufacture of fusion fuel targets — key steps in the company’s plans to turn laboratory science into grid-scale energy.

The ambition is taking physical shape. In July, Inertia opened a 50,000-square-foot headquarters in Livermore, California, with facilities for developing fuel-target manufacturing processes and laser technologies. The company also announced that it had recruited leaders from businesses, including Apple, Corning, and Waymo, to help build its manufacturing and supply-chain capabilities.

Want to learn more about raising major capital for your startup? Grab your Disrupt pass to lean in on Lawson’s onstage conversation. Save up to $200 before September 25 at 11:59 p.m. PT.

What you’ll leave with when you lean in on this session at Disrupt

Lawson’s experience spans more than one successful technology business. Alongside co-founding Twilio, he worked in early product leadership at Amazon Web Services and was the founding CTO of StubHub. His career brings a perspective shaped by building products, teams, and companies across different markets.

At TechCrunch Disrupt 2026, the focus turns to how that experience informs Inertia’s approach to talent, timelines, fundraising, and difficult engineering questions. It’s a conversation about applying hard-earned experience without assuming that the methods that worked in software will translate unchanged to fusion energy.

Image Credits:Steve Jennings / Getty Images

Consider the leadership questions underneath that transition. How do you build the right team when the expertise required extends far beyond your own? How do you set credible milestones while keeping an ambitious mission moving forward?

For founders, those questions offer a useful way into the discussion. For investors, they provide a lens through which to view the relationship between technical progress, capital, and commercial potential. Register here to save up to $200 on your pass before prices go up on September 25 at 11:59 p.m. PT.

Looking beyond the breakthrough headline at TechCrunch Disrupt 2026

The appeal of this session isn’t simply hearing a familiar founder explain an unexpected next move. It’s the opportunity to examine the work behind a company whose ambitions require scientific expertise and business execution to advance together.

Whether you’re building in climate technology, developing a hardware product, or leading a startup through unfamiliar territory, Lawson’s journey offers a different perspective on what it means to scale.

Make the conversation part of your TechCrunch Disrupt 2026 experience, then connect with founders, investors, and operators across the event’s stages, exhibition floor, Startup Battlefield, and networking opportunities.

Big ambitions make headlines. The decisions behind them build companies. Join Jeff Lawson at TechCrunch Disrupt 2026 and hear what comes next. Register here to join this session and 200+ other sessions led by 250+ tech leaders. Savings of up to $200 end on September 25 at 11:59 p.m. PT. Get up to 30% off on group passes of four or more.

Chinese AI startup Manus, which earlier this year had to break off a merger with Meta, is in discussions to raise $500 million at a $4 billion valuation now that it has resumed operations as an independent company, The Wall Street Journal reported, citing anonymous sources.

Potential investors in the round include IDG Capital, Boyu Capital, and battery maker Contemporary Amperex Technology, as well as existing backers Tencent, HSG, and ZhenFund, the Journal reported. Manus is also said to be considering a restructuring exercise to prepare for an IPO in Hong Kong.

Manus, which went viral following a demo of its AI agent last year, relocated its staff to Singapore in mid-2025 before announcing a $2 billion acquisition deal with Meta that December. The startup was said to be pulling in annual recurring revenue of over $100 million at the time.

However, intensifying worries in China over losing AI talent and researchers to the West culminated in Beijing blocking the deal, citing potential violations of export controls and foreign investment rules.

Manus has since been untangling itself from the American social media giant, and its early investors and backers have reportedly helped the company buy back its shares at a valuation of about $2 billion.

As part of its separation from Meta, the company told its users this August that they would have to export and back up their own data because it had to delete data generated following Meta’s acquisition to “comply with regulatory requirements in specific jurisdictions.”

The company this month said it has resumed independent operations, and that its founding team will continue to lead it.

Manus makes AI products and agents that are pretty similar to what companies like OpenAI, Lovable, and Replit offer. It offers a chatbot and vibe-coding tools to let users build apps and websites, create designs and presentations, generate video, and more.

Manus did not immediately return a request for comment.

Google is testing a new product designed to help families coordinate with the support of an AI agent. This week, the search giant introduced a new version of CC, an AI agent designed to work across email, calendar, chats, and tasks. With the update, CC now focuses on keeping families organized, planning for the day ahead, and even handling family-specific tasks, like signing permission slips, creating shopping lists, or crafting weekly meal plans.

The move to make CC more family-focused comes as a number of AI startups are experimenting with how to best fit agentic AI into consumers’ lives. Some AI tools, like Ollie and Fambot, for instance, have aimed their services directly at helping parents and families.

Google’s CC, meanwhile, began its life as a productivity agent that connected to Gmail, Google Calendar, Google Drive, and the wider web to understand your day, then deliver a “Your Day Ahead” briefing to your inbox. In May, the tool came to the Gemini app as “Daily Brief.”

Google said user requests showed that people wanted to use the feature more for household management tasks. They wanted AI to help them keep up with kids’ school schedules, sports practices, bills, meal plans, and more.

As a result, Google shifted CC to become an agent for families that helps run their households.

Image Credits:Google

Now, CC is getting its own Google account, so it can collaborate with family members on tasks, while also maintaining its own specific set of permissions for accessing data. Google explains that each family member chooses what they want to share — like emails about school events, sports, clubs, birthday party invites, doctor appointments, or anything else. These can be forwarded to CC via its email address, or shared automatically.

To automate sharing, users can pick the email senders whose messages they always want to share with CC going forward, like schools, travel companies, clubs, or sports teams. (CC will also suggest email senders to add on a weekly basis.)

Image Credits:Google

The company says CC currently supports up to six family members who can share information and collaborate with the agent.

Beyond email, CC can also coordinate household activities by tracking important dates and to-dos, and then automatically adding them to the calendar or a shared task list. That means, for instance, every time the orthodontist sends an email confirmation of your next appointment or the school announces a teacher workday, CC can put it on the shared calendar.

Image Credits:Google

What’s more, the agent can manage select tasks on a family’s behalf, like filling out permission slips or activity registration PDFs, making school supply shopping lists, planning weekly meals, figuring out drive times between activities, creating shared Docs or Sheets, and more. It will even ask for missing details, as needed, and update its group memory so it can be more helpful going forward.

Google notes that CC runs on its own isolated cloud computer, powered by Gemini and Google’s agentic harness, Antigravity. The experiment is only available for U.S. users with a personal Gmail account, who are ages 18 and up.

That’s a big drawback for the time being because it means older kids, like tweens and teens, can’t use the service unless they lie about their age on their Google accounts. It also means they can’t use it with their school-provided emails, despite the fact that many attend schools that run everything on Chromebooks and Google apps and have inboxes filled with school-related updates.

CC could help larger households, where parents have to coordinate with each other and with other adults, like grandparents, caregivers, nannies, or adult children still living at home, which tends to be more common these days.

Existing CC users will be invited via email in the coming days to upgrade their accounts, Google says, while new users can join a waitlist to access the AI agent.

隨著釣魚、假冒網站與網域挾持等威脅增加，台灣網路資訊中心（TWNIC）近日呼籲企業重新檢視網域與網站的基礎安全，並鼓勵企業利用TWNIC提供的網站健檢服務check.twnic.tw，進一步掌握自身網站與網域的防護狀況，特別是DNSSEC、RPKI及域名安全鎖（Registry Lock）等數位信任機制的部署情形。

TWNIC指出，企業過去談資安，多聚焦於防火牆、系統弱點與帳號安全，但隨著官方網站、電子郵件成為重要對外窗口，加上生成式AI降低假冒網站製作門檻，網域、DNS與路由等基礎設施也需要納入企業資安治理。

基於上述風險，TWNIC近年以「Trust in Every Click」為核心，持續推動臺灣網路基礎設施的安全與韌性。綜合TWNIC此次提出的建議，企業可優先關注3項較容易被忽略的防護重點，包括在DNS安全層面導入DNSSEC、在路由層級部署RPKI，以及在網域異動管理導入Registry Lock。

RPKI相對成熟、DNSSEC部署仍低

對於上述機制目前在國內的導入情況，我們進一步詢問TWNIC，並取得相關統計資料。其中，已推動多年的DNSSEC，整體部署比例仍偏低。

在DNSSEC方面，其運作方式是透過驗證DNS資料的來源與完整性，降低解析結果遭偽造或竄改的風險。過去多年來，網際網路名稱與號碼指配機構（ICANN）持續提倡擴大DNSSEC部署，呼籲尚未受到DNSSEC保護的域名全面導入，若未部署DNSSEC，一旦DNS資料遭到未經授權的修改，使用者可能在不知情的情況下被導向攻擊者控制的網站。DNSSEC雖無法防範所有DNS攻擊，但可透過數位簽章偵測DNS資料是否遭到竄改，降低DNS欺騙與錯誤導向的風險。

TWNIC提醒，過去DNSSEC導入門檻較高，但如今多數DNS代管服務業者已提供支援，通常只需一鍵即可啟用，大幅降低使用難度，但仍需要企業主動開啟這項功能，因此企業可重新評估導入可行性，並及早啟用。

儘管導入門檻已降低，目前DNSSEC的整體部署比例仍低。根據TWNIC提供的數據顯示，截至9月14日，以「.tw」或「.台灣」結尾的域名中，共有10,356個完成DNSSEC設定，占比為2.47%，仍有相當大的成長空間。TWNIC表示，後續將優先推動公部門及關鍵基礎設施導入，並發展自動化簽署及簡化啟用流程。

在RPKI方面，這項機制主要透過ROA（路由來源授權）聲明哪些ASN有權宣告特定IP位址區塊，再由ROV（路由來源驗證）查核收到的路由宣告。

目前臺灣ROA覆蓋率已達98.06%，相對成熟；而在ROV過濾率的部分，臺灣為55.98%，全球為27.08%，臺灣領先全球一倍。下一階段，TWNIC將把重點放在擴大ROV部署，持續與電信及ISP業者合作推動相關查核機制。

在Registry Lock方面，則是從註冊管理機構層級限制重要網域異動，透過人工複核與多重離線身分驗證，降低網域遭未經授權修改、移轉或刪除的風險。TWNIC自去年推出這項服務，目前約有24個核心域名啟用，後續推動將以關鍵資產覆蓋為重點，號召更多重要企業實施網域治理。

除了上述網域自身的安全機制，從企業對外品牌與商譽保護的角度來看，TWNIC另提醒企業留意網域仿冒風險，包括選擇容易辨識、與品牌具有明確關聯的網域，並持續監控是否出現高度近似的網域，以便及早發現潛在仿冒情形，降低品牌信任與商譽受損的風險。

網域可以委外管理，但控制權不應外包

另外，TWNIC也特別提醒企業注意網域委外管理的權責問題。對此議題，TWNIC執行長余若凡強調，即使將網域註冊、網站建置或DNS設定委由外部廠商處理，網域所有權、管理帳號與異動權限仍應掌握在企業手上。

網域及相關帳號也應登記在企業名下，避免合作關係改變後，面臨帳號難以取回或產生所有權認定問題。委外合約則應明確規範DNSSEC、SPF、DKIM、DMARC等安全設定，並要求網域或DNS異動須經企業同意，同時約定資安事件通報，以及合作終止後帳號、權限與相關設定的交還責任。

若要掌握自身網站與網域目前的安全狀態，TWNIC也提供免費的check.twnic.tw網站健檢服務。這項服務自2019年推出，並於2025年9月改版，目前可檢查HTTPS、DNSSEC、IPv6、RPKI、網站安全設定及Registry Lock等項目。根據TWNIC統計，自2025年8月至2026年9月10日，服務累計使用超過57萬次，其中今年6月至8月即達21萬次，反映網站基礎安全檢查的使用需求持續增加。