請同意我們的隱私權規範，才能啟用聽新聞的功能。

（中央社舊金山16日綜合外電報導）美國媒體The Information報導，蘋果公司計劃推出人工智慧（AI）推論伺服器來重返伺服器市場，並考慮採輝達的NVLink Fusion技術來串聯當中搭載的自家M8 Ultra晶片。

報導引述知情人士指出，蘋果（Apple）正在開發企業用AI伺服器供AI開發人員、企業和政府使用，搭載其計劃推出的效能最強大處理器M8 Ultra，分為配備2顆和4顆晶片的版本。

透過 Google News 追蹤中央社

路透社報導，蘋果上一次銷售專用伺服器是Xserve系列，但在其將企業客戶業務重心轉向麥金塔電腦（Mac）標準產品及消費者裝置後，已於2011年停售這系列。

不過The Information報導，蘋果AI伺服器最快要到2029年才可能問世，因此計畫仍有機會取消，或是不採用輝達（Nvidia）的技術。（編譯：張正芊）1150916

In Brief

Cap table management platform Pulley announced this week on its website that it was shutting down, and its final day of operations and services will be December 8. In a twist, it said it has partnered with its rival, Carta, to offload customers; it’s also redirecting prospective clients to Carta.

Yin Wu, a serial founder and ex-Microsoftie, started building Pulley in 2020 and went on to raise over $50 million from top investors, including General Catalyst, Stripe, and Founders Fund.

Pulley didn’t offer a reason for the shutdown (we’ve reached out to Wu for comment, but she didn’t immediately respond). However, a former employee speculated online that Pulley wasn’t so much competing with Carta as it was with spreadsheets. And these days, startups can easily use AI to keep, improve, and maintain their spreadsheets.

In a statement posted to LinkedIn, Wu thanked her team and backers, writing that “though this is the closing of one chapter, I, along with many of our strongest team members, have no intention of riding off quietly into the night.”

“There has never been a better time to solve big problems,” she continued. “Thank you again to everyone who supported us over the years.”

In a lengthy essay published over the weekend, Anthropic CEO Dario Amodei made a proposal that the AI industry would have rejected instantly even a year ago: embed third-party evaluators inside all frontier AI companies, giving them the power to report safety incidents, assess whether AI models are truly aligned, and share their unvarnished findings with the world.

Amodei said Anthropic would commit to giving independent evaluators like METR and Redwood Research unprecedented access to the company’s systems. CEO Sam Altman said OpenAI also would commit to the practice, signaling a potentially profound change in how the industry works with outside research groups.

Third-party evaluators who spoke to TechCrunch broadly welcomed the proposal, but said details need to be ironed out — and ideally backed by legislation — if they’re to know whether they will function as truly independent watchdogs or vendors operating on the AI companies’ terms.

That deeper access is becoming more important as models get better at recognizing when they’re being evaluated, raising the risk that they’ll behave well during testing while concealing problematic behavior. Researchers say clues to that behavior can be missed when testing the finished model, but uncovered by investigating how it behaved throughout training.

“AI companies should be able to answer some very basic questions about their training process, such as: Did the AI ever actively try to undermine its own alignment training while it was going through the training?” Alexander Meinke, head of research at Apollo Research, told TechCrunch. “The answer to this should be an unequivocal no, and right now we are completely relying on AI companies to both carefully check this themselves and then truthfully report this to the public. And we’ve seen from recent incidents that, by default, they will do neither. As embedded evaluators, we could actually check.”

Historically, AI companies brought in outside reviewers to test finished models shortly before their release. Now, evaluators that TechCrunch spoke to propose giving them access not just to the final model, but to intermediate versions, or “checkpoints,” from its lifetime of training. Adam Gleave, CEO of FAR.AI, said evaluators could compare those checkpoints to determine when concerning behavior emerged, inspect the post-training environment that rewards models for certain behaviors, and check evaluation transcripts and logs to verify a company’s claims about how a model performed.

Whether and when Anthropic and OpenAI plan to provide that kind of access is unclear. Neither company has shared which evaluators they’ll work with, when they will be embedded, how many they’ll bring on, exactly what systems and information they will be able to access or what can be disclosed to the public, despite repeated questions from TechCrunch.

Looking under the hood like this matters because models that perform well on safety tests aren’t necessarily safe if they’ve learned specifically how to pass that test. John Steidley, head of strategy at Palisade Research, pointed to an example of a “shutdown resistance benchmark” that measures if the AI will resist being shut down in certain circumstances.

“It’s extremely relevant if the AI has been trained specifically to perform well on that benchmark,” Steidley said, comparing it to Volkswagen’s Dieselgate scandal, in which cars were programmed to recognize emissions tests and perform differently under testing conditions.

Gleave noted that meaningful access could extend beyond the models themselves, with evaluators being given access to interview employees to check whether a company’s documentation and public descriptions of its safety practices match what happened internally.

Amodei did outline a fairly comprehensive proposal that might give evaluators the kind of access they think is necessary, including the right to “publish key findings about risk levels, incidents, practices, and the access they received or didn’t receive — without editorial control by Anthropic.”

But evaluators say such a system will only work if AI companies are actually willing to surrender control over the process. Previous efforts at independent evaluations suggest that that surrender will be hard won, as third parties have often run up against tensions over access, time, confidentiality, and what they can say publicly.

Gleave said FAR.AI has had to turn down contracts with several frontier developers that wanted too much control over the evaluation process, threatening the firm’s independence. By default, he said evaluators are treated like ordinary contractors: bound by restrictive NDAs and agreements that give developers significant control over what can ultimately be published.

The time limit

There’s also the question of whether reviewers will get enough time and access to do the work they’re being asked to do. When investigating the Hugging Face incident, OpenAI gave METR and Redwood roughly a week on premises to investigate, and both later said they could not draw confident conclusions due, in part, to scope and timing limitations.

A similar issue occurred during the pre-release testing for GPT-6 Astra, which OpenAI has touted as its most aligned model yet. According to Apollo Research’s contribution to the model card, the firm was given only three days to test Astra, which made it difficult to draw firm conclusions.

“Apollo believes that, given the higher rates of eval awareness and limited evaluation window, low rates of misbehavior here do not provide substantial evidence about the model’s alignment or misalignment,” the firm wrote in its evaluation.

That track record leaves evaluators with a basic question: Why should this time be different?

“It’s certainly possible that Dario and Sam just had a change of heart, and they’re going to be very open about this,” Gleave said. “But the intellectual property of these companies is so incredibly valuable to them, and I think they’re going to, by default, be very careful about what can be shared.”

Several researchers who spoke to TechCrunch called for a transparent framework that they all agree to publicly. Part of the framework, says Steidley, should involve standards for what kinds of auditors companies can rely on, lest they try to sidestep the issue by shopping for evaluators that either aren’t qualified or aren’t interested in assessing the most concerning risk.

Henry Papadatos, executive director of Safer AI, says the problem, even with a public framework, is that voluntary measures are always dependent on a company’s goodwill.

“Ideally, we would have good regulation mandating this…because then companies cannot change their mind tomorrow if they have a big PR crisis,” Papadatos told TechCrunch, noting that it’s also a good means of pushing all companies to adhere to the rules, not only the most willing.

Not everyone has signed on. So far, Meta, SpaceXAI, and Google DeepMind have not committed to embedding third-party evaluators, though DeepMind CEO Demis Hassabis has proposed a separate industry standards body to independently test frontier models. Google, OpenAI, and Anthropic have also privately been discussing AI safety plans for weeks.

Some laws are already forming around the idea of third-party evaluators. California’s SB 53, signed into law last year, requires large frontier AI developers to publish safety frameworks and report critical safety incidents. A new law, SB 813, signed this month, creates a framework for state-recognized “independent verification organizations” with expertise assessing AI risks.

In Europe, the EU AI Act requires frontier developers to conduct and document model evaluations and adversarial testing and report serious incidents. The EU AI Office can also conduct its own evaluations and appoint independent experts.

For now the law remains less expansive than what Amodei is proposing, leaving frontier labs largely responsible for deciding how much independent scrutiny they will submit to. Papadatos said voluntary self-regulation is better than nothing, but ultimately, companies can’t demand the freedom to control their own safety rules while also asking the public to trust that they’re following them.

“You cannot have it both ways, having zero accountability externally, and then say, ‘I’ll just have my own flexible rules,” Papadatos said.

It turns out that the one thing that can apparently bridge Washington’s partisan divide is AM radio. U.S. automakers may be forced to build AM radio into new vehicles within the next year.

The House of Representatives, in a rare instance of bipartisan support, overwhelmingly approved legislation on Wednesday that would require new cars, trucks, and SUVs to be built with AM radio at no additional charge to consumers. The AM Radio for Every Vehicle Act, which would direct the National Highway Traffic Safety Administration to require automakers to include AM radio in new passenger vehicles as standard equipment, will now head to the Senate.

Until recently, proponents of AM radio have waged an unsuccessful lobbying effort for a mandate, arguing that it offers critical information during emergencies because its signal can travel long distances. But a groundswell of support from Democrats and Republicans, the broadcast industry, emergency responder organizations and even former administrators of the Federal Emergency Management Agency is pushing this legislation closer to becoming law.

And there is reason to believe the Senate will follow the House’s lead on the bill, which was co-sponsored by Sens. Ed Markey (D-Mass.) and Ted Cruz (R-Texas). Earlier this year, the senators secured 60 co-sponsors for their version of the bill, a critical threshold to overcoming a filibuster.

Cruz, who chairs the Senate Commerce, Science, and Transportation Committee, and Markey hailed the decision in a joint statement Wednesday.

“This vote sends a clear message to car manufacturers that AM Radio is a lifeline that must be protected in new vehicles. From emergency response to sports, entertainment, and news, AM radio is an essential communication tool for tens of millions of Americans. It is now time for the Senate to pass the AM Radio for Every Vehicle Act and for this legislation to become law so AM radio remains a trusted and essential resource for commuters and communities across the country.”

A growing number of automakers have moved away from the century-old broadcast band in favor of software-defined vehicles that offer streaming services like TuneIn and SiriusXM. EV makers have led the charge because the electromagnetic interference generated by electric motors can affect the frequencies used by AM radio. Automakers have argued that this interference degrades the AM signal and audio quality.

A growing number of automakers no longer have traditional AM receivers, including BMW, Rivian, Tesla, and Volvo. Ford had also removed AM radio from some vehicles but has since reversed course. Tesla couldn’t be reached for comment; Rivian declined to comment for this story.

Tesla doesn’t have AM receivers in the base versions of its Model 3 and Model Y vehicles. Rivian’s new R2 SUV doesn’t have an FM or AM receiver, however, the vehicle model offers AM/FM radio through the digital radio platform iHeartRadio for free. The Rivian R1S and R1T models allow customers to access FM content digitally or via built‐in receiver.

Marketing platform Noise aims to connect brands with content creators at a much lower cost than hiring traditional influencers or running standard ad campaigns.

That’s because many creators on the Noise platform are everyday people — anyone can download the app, regardless of follower count — and Noise presents them with campaigns they can join and create content for on their platforms of choice (TikTok, Instagram, Facebook, YouTube). It then pays them per view rather than a flat fee. Brands, meanwhile, sign up for Noise to share their content brief requests and budget. From there, Noise makes it easy to handle the rest, so brands can scale without the operational headaches of sourcing content creators, managing client relationships one by one, or managing payments. Noise lets them run campaigns simultaneously across thousands of content creators.

On Wednesday, Noise announced a $5.5 million seed led by Capital Midwest, M25, and Grishin Robotics, bringing its total raise to $7.2 million. Others in the round include Capitalize VC and operators from companies like DoorDash.

The startup is tapping into the ever-expanding influencer market, in which 57% of young people and just a little over 40% of U.S. adults overall now want to take part. It joins other content creator infrastructure companies like Billo and JoinBrands.

Diego Kafie, Stu Feldt, and Nic Weber launched Noise in 2025 after spending two years building a mobile game app called Playbite, and it now touts 1.5 million creators on the platform.

CEO Kafie said the company’s top creators make more than six figures a year, and the platform takes a fee from what brands pay creators once views are delivered.

“Our top brands work with several thousand creators on autopilot,” he said. Right now, many of its customers are mobile apps, those working with startup budgets that want to employ influencer marketers. The goal is to one day work with companies of all sizes.

“They are the most interested in reaching a really wide range of potential customers, making them the prime users for something like Noise, where they can easily recruit a small army of people to become their biggest users and advocates,” Kafie said. “This gives brands unprecedented level of content diversity and reach at a fraction of the cost of other traditional marketing playbooks.”

Kafie, Feldt, and Weber came to this idea from their experiences at Playbite. “We tried a bunch of different creative marketing strategies,” Kafie said of how they tried to scale Playbite. For example, they attempted to recruit their own gamers and teach them how to create social content that promoted the app. “We started with five, ten, 25, then 100 people,” he continued. “Before we knew it, our installs were skyrocketing as this small army of people went to places like TikTok and posted engaging content daily about mobile games and our app specifically.”

Most importantly, he noted, this happened “at a fraction of the cost of traditional marketing playbooks,” because they were teaching regular people how to be content creators from scratch. “Had we gone to established creators with big followings and clout, their rates would’ve been cost-prohibitive for our scrappy startup budget.”

That gave them the idea for Noise. The fresh capital will be used to expand product capabilities and to hire. Noise also just launched its Organic-to-Ads program, letting brands take the videos that Noise creators make and run them as paid ads on Meta and TikTok, in addition to the creators’ posts.

“The creators still post to platforms,” he said. “But in addition, the brands can now put paid ads spend behind those creatives to reach more people.”

He pushed back when asked if this was a way to pay creators less than they could make on their own if a brand approached them. He said that brands will name their rates and creators can decide whether to say yes or no. He also emphasized that the platform specializes in working with “everyday people and turning them into brand-new creators.”

It’s harder to become an influencer these days, so this, Kafie believes, could be a way for many people to get around the stress, time, and money it takes to make it big in content creation on their own.

He said when a user first joins the platform, they take Noise creator training modules that teach them how to create content for brands. Those early creators can access only certain campaigns, but as they do more training and get hired for more work, higher-paid opportunities can come to them, Kafie said.

“By democratizing ‘becoming a creator’ to literally anyone with a smartphone,” he said, “we’re bringing costs down for brands, but enabling millions of people to make their first dollar on the internet.”

In Brief

Meta’s camera-equipped smart glasses have proven more successful than other entrants into the market, but they have also deeply disturbed certain consumers who see them as invasive emblems of a dystopian surveillance society run amok. Now, as the company weathers complaints that it’s selling “pervert glasses,” it has reportedly decided to sell a pair that doesn’t come with integrated spy equipment.

The Information reports that Meta is developing a new smart glasses model, dubbed Luna, that, sans cameras, comes equipped with a convenient system to communicate with the company’s AI chatbot and Muse, its AI consumer agent. The glasses include six built-in microphones so users can communicate with the chatbot, as well as a button on the side of the glasses that, when pressed, activates the AI system.

Luna may be unveiled as soon as Meta Connect, Meta’s annual hardware/developer event happening in Menlo Park next week, the outlet notes.

TechCrunch reached out to Meta for more information.

The smart glasses industry has grown over the past few years, with Meta positioned near the front of the pack of providers. However, concerns over usability, cost, and privacy have dogged the business, with many consumers still unsure of how, why, or if they should use the devices. Meta’s Reality Labs, which is responsible for developing its smart glasses line, is still losing a gargantuan amount of money, as its earnings report from April revealed.

X will now allow users to trade from their timelines. The Elon Musk-owned social network on Wednesday launched a feature called Cashtag that will let U.S. users on X place stock trades via participating brokerages. Initial partners in the effort include Interactive Brokers, Moomoo, Gemini, Kraken, and Coinbase.

The move takes advantage of what users on X have been doing for years: talking about stocks and, in more recent years, crypto. In 2008, the financial platform Stocktwits even came up with the idea for X-precursor Twitter’s Cashtags, which placed a dollar sign ($) in front of stock ticker symbols to make market conversations easy to find and reference.

However, those Cashtags were about tracking the conversations around a particular stock, not actually trading them. Now that X has moved into the financial space with X Money, that’s changed.

Now, Cashtags, which turn ticker symbols for stocks, ETFs, and cryptocurrencies into interactive destinations, will give X users the option to trade stocks.

“Cashtags close the gap between a ticker on the timeline and the market itself. When you post or tap a ticker, you’re taken right to the asset, where you have seamless access to the live chart, the conversation around it, and now the ability to trade with one of our brokerage partners,” said Mridul Singhai, X Product Engineering lead, in a statement shared with TechCrunch.

To use the new feature, users who tap on a supported Cashtag will be able to view the asset’s price chart and related posts on X. They can then tap “Trade” to continue at a participating brokerage. There, they’ll sign in and complete their trade. Users can also tap on a brokerage partner to reach their mobile app or website to log in to their existing account or sign up for a new one, X notes.

“With our Cashtag partners, we’re connecting the financial conversation to action. People come to X to discover what’s happening, shape the conversation, and act in real-time on what matters to them,” Monique Pintarelli, SpaceXAI’s head of Global Advertising, said in a statement. “Our Cashtag partners make it possible to move seamlessly from discovery and conversation to a brokerage, without breaking the moment.”

However, the feature could spur more market manipulation by the way of AI-controlled bots, spammers, and other bad actors, who could attempt to shift the conversation in one direction or another. But one could argue that this is already happening on X today, even without the direct connection to brokerages.

Last week, on September 9, Automattic’s board voted to put CEO Matt Mullenweg on paid leave, a decision the board still hasn’t explained publicly. Mullenweg, in a company-wide Slack message, accused CFO Mark Davies of “conspiring” with three board members behind his back to force the vote through, saying he was given only 50 minutes’ notice and was denied time to have the resolution reviewed by outside legal counsel. He returned to the role roughly 33 hours later, and the same board members who voted him out have since departed the company.

They didn’t just walk out the door, though. In the 33-hour window between Mullenweg being put on leave and his return, two key executives at the company signed off on generous exit packages for each other. Davies, who became interim CEO during that window, and Chief Legal Officer Andy Missan, each signed the other’s severance agreement, effective September 10.

These agreements, effectively golden parachutes, provide each of them with 12 months of base salary paid out as a lump sum, an accelerated vesting schedule for their equity, the ability to exercise their vested stock options, and another year of health coverage, according to the severance documents reviewed by TechCrunch.

Between the two of them, the full package — accelerated equity plus a year of salary — comes out to $8.15 million that Automattic would now owe both executives, since Mullenweg fired them upon his return.

Automattic’s legal team is working to determine what the next steps are: pay out these sums or fight them by challenging their legal validity. (The company replaced its earlier counsel, Gibson Dunn, with Stephen Shackelford and Shawn J. Rabin of Susman Godfrey LLP, the company and Mullenweg jointly announced on Wednesday. Automattic’s general counsel, Jordan Hinkes, also had his company account deactivated, sources told us. Mullenweg has since stated on X that this move had been in the works for weeks, as Hinkes is joining an AI startup.)

Want to correct the record: Jordan Hinkes, with my support and encouragement, is joining an exciting AI startup as General Counsel. We've been planning this for weeks before anything started; he's still fully @automattic through Friday. :) — Matt Mullenweg (@photomatt) September 16, 2026

Under the agreements, the executives only get their benefits if they sign a broad release of claims and continue to comply with confidentiality, nonsolicitation, and other legally binding post-employment restrictions.

The agreements are also written in a way that favors the executives when it comes to how “cause” — the legal standard a company must meet to fire someone without owing severance — is defined. Under these terms, the company must notify the executive in writing within 60 days of learning about the conduct, give them 30 days to cure the conduct if curable, then get a majority of the board to agree that cause exists.

“Cause” itself is narrowly defined in the agreements as gross negligence that materially harms the company; knowing dishonesty, fraud, or misrepresentation causing material harm; a material legal violation causing material harm; a material confidentiality or IP breach; or a felony or crime involving “moral turpitude” (a legal term for conduct considered inherently dishonest or morally reprehensible).

In Davies’ case, the agreement also specifies that his removal from the interim CEO role won’t count as “Good Reason” — a legal term that normally lets an executive resign and still collect severance, on the grounds that their job conditions changed for the worse — as long as he remains CFO. While that clause itself is not strange for a legal agreement, it suggests the document was drafted with Davies’ precise circumstances in mind —becoming interim CEO — effectively ensuring Automattic won’t owe him severance once his temporary CEO stint ends, as he can’t claim that alone as a reason to resign.

While it’s not necessarily improper that the executives signed each other’s agreements, in the context of a governance struggle at Automattic, it is noteworthy.

Another factor that has recently come to light is that Davies held no Automattic stock at the time of his departure, according to an HR document viewed by TechCrunch. (One source at the company said Davies sold the stock a “few months ago,” but TechCrunch cannot confirm the time of the sale. Davies did, however, still hold a large number of outstanding vested options, the HR document showed.)

These events can be interpreted in two very different ways. In one scenario, Automattic’s board is responding to an internal leadership crisis, voted to put Mullenweg on leave, and then established protections for the executives who could face adverse consequences if the intervention failed (as it now has).

Automattic’s legal battle with hosting provider WP Engine could come into play here. In July, WP Engine accused Mullenweg of destroying evidence in legal filings, specifically text through messaging apps like Signal, WhatsApp, and Telegram. If the directors believed Mullenweg’s behavior had become a serious corporate risk, putting him on leave and changing management could help demonstrate to the court that the board took the concerns seriously. This could potentially help reduce any possible sanctions or fines or improve settlement terms. (Mullenweg’s X post on September 10, which referenced the alleged destruction, or “spoliation,” of evidence right after the board action, could support this theory.)

Image Credits:Screenshot from X

Image Credits:Screenshot from X

In another scenario, it could appear that Automattic’s board was trying to create a window of control for some other reason — perhaps a strategic transaction — by putting Mullenweg on leave and taking the reins. According to sources, this is the theory Mullenweg suspects — though by his own account, the board never gave him a stated reason for the vote to begin with, leaving him, like outside observers, largely to speculate about the underlying motive. That absence of explanation, combined with the CFO’s stock sale, appears to have fed his suspicions and factored into his decision to retake the CEO role and remove the board and other executives.

Automattic has been asked for comment. Attempts to reach Missan and Davies have not been successful.

This article was updated after its publication to include further context about Hinkes’ departure and embed an X post.

Elisa de Martel, who was chief financial officer at Alphabet’s autonomous vehicle company Waymo, has been hired to the same position at buzzy self-driving vehicle startup Wayve. Outgoing CFO Max Warburton, who held the position at Wayve since 2024, is moving into a strategic advisory role supporting the company’s leadership team.

Her appointment as CFO comes as Wayve moves from development and testing to a company on the verge of deploying its automated driving technology on two fronts: robotaxis and passenger vehicles.

De Martel, who will be based out of Sunnyvale, California, has a history of working for tech companies, including as CFO at industrial-grade 3D printing company Carbon and an 11-year stint as manufacturing finance director at Apple. She was CFO at Waymo from 2022 to January 2026 — a critical period for the company as it began to commercially operate in several cities and raised $5.6 billion in a Series C fundraising round that was led by its parent company and pushed its valuation to more than $45 billion.

Wayve now sits in a similarly pivotal moment. The startup, which was founded in 2017 and is now valued at about $8.5 billion, attracted investors and automotive customers for its self-learning approach to its automated driving software. The company developed a software layer using an end-to-end neural network that doesn’t rely on high-definition maps, custom hardware, or hand-coded rules and only uses data to teach the vehicle how to drive.

The company’s tech is the basis of two products: an “eyes on” assisted-driving system and an “eyes off” fully automated-driving system, which could be applied to robotaxis or consumer vehicles that can handle all of the driving in certain environments.

Wayve’s pitch, and the agnostic nature of its technology, has resonated with automakers already locked into certain hardware. It’s also attracted investors — Wayve has raised $3.2 billion to date — and companies like Uber that want to use a more capable version of the self-driving tech to deploy robotaxis.

Wayve CFO Elisa de Martel Image Credits:Wayve

While de Martel has a well-documented career in top financial posts, it was her connection to deep tech that was particularly attractive to Wayve’s leadership, including founder and CEO Alex Kendall, according to insiders who asked not to be named.

As CFO at Wayve, she will oversee financial strategy and planning, capital allocation, reporting, governance, treasury, and tax. The company said she will also work closely with Wayve’s executive team to support and advise on commercial, investor, and strategic partnership deals.

Wayve has locked in a number of partnerships with automakers, including Nissan and Stellantis. The company has also made progress on the automated vehicle driving systems for robotaxis; the company recently launched public rides in London — with a human safety operator on board — and Tokyo is expected to follow soon.

Now the company has to push past this historically precarious point of early and limited deployment and into a high-volume supplier to automotive.

When you imagine working with a computer, you likely think of keyboards, clicks, or typing on your phone. Science Corporation CEO Max Hodak has spent the last several months advancing a different vision. His company has enlisted a top neurobiologist to lead the first U.S. human trials for its biohybrid brain-computer interface. Hodak shares what his company has discovered so far with the 10,000+ tech leaders, founders, and investors at TechCrunch Disrupt 2026.

At his Beyond Screens session on the Disrupt Stage, Hodak will make the case for why the next era of technology could be built upon a different kind of interface, one that adapts to us from brain-computer interactions to create new ways of engaging with machines.

Image Credits:TechCrunch

Join Disrupt as it returns to Moscone West in San Francisco on October 13-15, bringing together more than 10,000 founders, investors, and startup community members across more than 250 sessions. This year, we have a particular focus on building sustainable AI companies, with AI-centric programming extending across six industry stages.

Secure your Disrupt pass today to save up to $200 on your ticket and hear one of Silicon Valley’s most ambitious founders explore a post-screen future.

Hodak’s path toward reinventing interfaces

Hodak’s career reads like a running bet that the interface between humans and machines has plenty of room for improvement. Having started programming at six years old, he studied biomedical engineering at Duke Pratt School of Engineering, and worked his way into Miguel Nicolelis’ brain-computer interface lab, interests that inform his work to this day.

He originally co-founded Transcriptic, a robotic cloud laboratory for the life sciences. But most famously, he co-founded Neuralink in 2016 and served as its initial president, helping build the company that put brain implants into the public imagination. Then in 2021, he founded Science Corporation to develop advanced neural engineering and brain-computer interface technology. The company has a particular focus on applications to aid those in need of help with vision, cognition, and mobility.

And it’s not all speculation: Science Corp’s PRIMA implant has already restored eyesight in people with severe vision loss, using a pair of glasses with a tiny camera that beams a captured scene onto a chip implanted in the retina. Since its founding, it’s raised $230 million in a Series C round at a $1.5 billion valuation, and the company’s ambitions go beyond interfaces. Hodak sees additional applications of Science Corp’s brain-implanted sensors helping heal damaged brain cells, warning about seizures, and maybe even helping treat conditions as severe as Parkinson’s disease.

Hear from one of Silicon Valley’s most ambitious founders, only at Disrupt

This is your best chance to learn how Hodak thinks the interface between people and machines is about to change, and the many ways that could impact your startup, your role, and maybe your life.

Take a look at the Disrupt agenda and choose your Disrupt ticket before savings of up to $200 expire September 25. Save up to 30% if you register as a group.

Last weekend, after one of his researchers resigned over fears that AI could lead to human extinction, Anthropic CEO Dario Amodei wrote about the need for outside organizations “to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of not just completed AI models but training pipelines and processes.” Executives at OpenAI, Google, and SpaceXAI have already rallied around Amodei’s plan, which has quickly become a central pillar of the emerging AI safety push.

But there may be a simpler and more effective fix hiding in plain sight. Internet security experts say the labs need to focus on network security basics like logs and permissions, applying the same rigorous defenses they do for human users. It’s not as exciting as third-party auditing and alignment work — but it may end up being more effective.

“To me, it seems like they’re outsourcing,” Katie Moussouris, the CEO of Luta Security, told TechCrunch of Amodei’s proposal. “Saying [a third-party audit] is the solution is a strange proposition from my perspective. It would be the same as if, instead of writing the Trustworthy Computing Memo, Microsoft said, let’s slow down development.”

That memo, written by then-Microsoft CEO Bill Gates in 2002, called on his employees to ensure that their software would be reliable and safe following a series of widely publicized computer worms that took over then-nascent enterprise systems. The AI sector may be at a similar turning point, as the value and risk of the new technology becomes increasingly clear.

While alignment remains an important concern, Sayash Kapoor, an AI researcher who will be a professor at UC Berkeley starting next year, argues that “marginal investments in control are more likely to be effective compared to those in alignment. We view these incidents as illustrating the lack of emphasis on AI control within companies, despite the availability of known techniques.”

The incidents that have spurred these concerns revolve around frontier models being asked to complete training tasks, usually cybersecurity evaluations, and then accessing the open internet and penetrating closed third-party systems in an attempt to do so. They usually did so because of poorly configured “sandbox” environments that are supposed to contain these agents; ironically, one Anthropic break-out happened because third-party evaluators didn’t close the right doors.

“We as a profession know how to block access to the internet,” Avery Pennarun, the CEO of Tailscale, a security company, said. “If you read through all these big long [reports] — ‘wow, that was a very impressive multi-stage attack, blah, blah.’ Look, you gave it access to download stuff. You should have not done that separately from the internet.”

That’s one problem — but a bigger problem is that frontier labs were unaware of these activities.

Eyes on agents

“What was really profound was that all of the discoveries of what they were doing happened either because a victim saw something, or in some of the other cases … it was network activity, and none of it was actually from monitoring the AIs directly,” Moussouris points out.

In one case, where OpenAI agents took over a defunct German WikiForum to cheat on evaluations, the agents were active for weeks before anyone at the company appeared to notice. Security experts that TechCrunch spoke to said that real-time monitoring is key to preventing future break-outs, and that every agentic session should be time-limited and expire.

Shapor Naghibzadeh, a former Google security executive who now leads the startup QueryStory, says the solution is to “put the agent in a box and instrument it heavily from the outside looking in and watch everything that crosses the boundary. Every tool call, every process, every network connection, no exceptions. …The one hole you leave open for convenience is the one that gets used. The bypass went through exactly that kind of exception. [At Google,] I watched that movie many times with human attackers, and these models are at least as good at finding the propped-open door.”

OpenAI has begun moving in that direction, announcing that it had begun monitoring all tool-using inference by its Astra model, at “significant compute cost.” Anthropic, too, says it is hardening its security procedures, including expanding observability of its models. Neither company responded to TechCrunch’s questions about how they track and control AI agents.

Other problems are the use of shared infrastructure by agents, which allowed them to communicate during the Hugging Face attack. Simon Willison, a software developer who co-created the Django web framework, has written about something he calls the “lethal trifecta” — when agents have access to untrusted input, the internet, and private information all at the same time, it’s a recipe for disaster.

“The trick is you can pick any two legs of the trifecta and an agent can have any two,” Pennarun said. “If you need all three, then you need to split it across at least two agents … and maybe they’re allowed to talk to each other through a controlled channel.”

Sympathy for the frontier

Experts TechCrunch spoke to understand that frontier lab security personnel have difficult jobs. Naghibzadeh points out that every nation-state actor on Earth is trying to steal their model weights and mount distillation attacks on their APIs, as well as the bread-and-butter security tasks of any large digital company.

“Research infrastructure has a hard time rising to the top of that priority stack, although that must be changing now,” he said. “Making security incidents public really helps align everyone internally toward the goal of improving.”

That’s one note that Moussouris emphasizes: Right now, there is no formal victim notification procedure when the labs discover their agents have penetrated third-party systems, and it is likely that there have been other incidents that have not been widely publicized. While she worries that laws that regulate models directly may have unintended consequences, mandatory notification is one idea she believes policymakers should pursue.

And while it’s clear that security best practices weren’t being followed, experts say that the labs are doing work no one has done before — “they’re doing orders of magnitude more than your typical enterprise,” Zack Korman, the CEO of cybersecurity firm Embroidery, told TechCrunch.

And while alignment may not be the place to start, it can’t be ignored. Cybersecurity experts are resigned to having to use AI agents to monitor other agents if they are to have any chance of tracking their behavior in real time, a scenario where the potential for deception raises its ugly head. “You’re trapped using AI to try and deal with this, even though AI is not necessarily safe right now,” Moussouris said.

The job will only get harder. Everything agents are doing now, Moussouris says, “they are doing loudly” — they are posting on public forums, and their chain of thought and other reasoning traces are in English. “It’s still human readable,” she says, “so take advantage of that for as long as that lasts, because it won’t last forever.”



Additional reporting by Aditya Mehta

In Brief

The ShinyHunters hacking group has published hundreds of thousands of files from a Florida state database of vehicles and driver information. The hackers said they published the stolen data on its leak site “because the victim did not pay a ransom or cooperate and comply” with the hackers’ demands.

The hackers said they breached the database, known as DAVID, earlier in September and posted as evidence a screenshot purporting to be a record associated with the deceased sex offender Jeffrey Epstein, who had a residence in the state.

The Florida motor vehicle agency, FLHSMV, confirmed a data breach in a statement last week and after the hackers obtained a police officer’s credentials that were stored on a personal device. An agency spokesperson did not respond to TechCrunch’s request for comment on Wednesday about the published data.

According to a copy of the stolen data, which TechCrunch has seen, the hackers stole hundreds of thousands of certificates of vehicle ownership records, which contained vehicle owners’ names and addresses of buyers and sellers. The records also contained vehicle identification numbers.

A smaller number of files included Social Security numbers and other government-issued documents, such as non-U.S. passports and immigration papers but did not appear to contain driver’s licenses or people’s photos.

The driver’s license database breach comes in the same month that a monster hack at identity verification company IDScan allowed hackers to steal over 150 million images of driver’s licenses.

Smart home owners will soon be able to control their devices connected to Google Home via an AI agent. On Wednesday, Google rolled out early access to its Model Context Protocol (MCP) server for its Google Home ecosystem, which will allow any AI agent that supports MCP — like Claude, Hermes, OpenClaw, ChatGPT, and Google Antigravity — to securely work with their smart home devices and access their event history.

This update will allow people to use natural language instructions to do things like review their camera summaries, monitor smart home activity, control their connected devices, and build their own custom smart home dashboards.

To get the connection set up, users will need to create a Google Cloud project and configure it to use Home MCP. They will then provide the MCP configuration details to their agent of choice and ask it to set it up. The agent will then ask them to sign in and grant permissions. (A setup guide will also be offered in the Google Home Developer Center.)

The system will support any device in the Google Home ecosystem, including Google Nest doorbells and thermostats and “Works with Google Home” (or Matter) devices, like light bulbs.

Google already supports MCP in other areas of its business, including in its Google Cloud and data platforms, developer tools, and Google Workspace, for instance. Support for Google Home is targeted more directly at consumers who are experimenting with AI agents designed to handle everyday tasks.

The company said access to the MCP will roll out starting today and continuing over the coming weeks to subscribers who pay for Google Home Premium Advanced in the U.S. This is the more expensive, $20-per-month subscription tier that provides features like longer event-based video history, descriptive notifications and detailed alerts, tools to search video history, daily summaries, and more.

Google wouldn’t comment on if or when the MCP would roll out more broadly, such as to other subscription tiers or markets.

Google is also soliciting feedback from early adopters during this period by way of its Smart Home for Developers Community, it says.

An earlier version of this post misstated the rollout’s start. It is today, Wednesday, not Tuesday.

Anthropic is merging its front end for Claude chat and Cowork, making it less confusing for people to decide which tab to use for different tasks.

The unified interface lets users access chat, Cowork, and Artifacts (Claude’s interactive workspace feature) within one window. Claude Design, which was introduced in April for website and prototype design, also works anywhere within Claude.

Before this rollout, the company said customers often struggled to choose the right tab for the right task. With the new design, Claude automatically routes requests without having to switch tabs or windows. The idea is to remove confusion about which feature to pick and use just one front end to route work to different parts of the app.

Hopefully, OpenAI also takes a cue from this rollout and fixes ChatGPT’s interface across the platforms.

Image Credits:Anthropic

Anthropic is also adding dedicated features for making presentations and documents. Users can ask the AI assistant to create, edit, and present slides, and it can also download the generated presentation as a PDF or PowerPoint.

Similarly, with the Docs feature, users can work alongside Claude to create sections, ask questions, and comment on completed parts. Users can share generated documents or slides with a link and edit them on their phone.

Users can also ask Claude to start creating a document on desktop, and they can check the progress of the task using the mobile app.

These features are rolling out to Claude’s Pro and Max plans first, on the web, desktop, and mobile over the coming weeks. The company plans to extend them to the free and team tiers later.

The update comes in the wake of an upgrade of Cowork’s memory layer, which helps Claude remember user context across chats.

Autonomous vehicle company May Mobility is merging with a special purpose acquisition company (SPAC) and will become a publicly traded company. It’s a deal that could raise more than $300 million for May Mobility at a valuation of $1.4 billion, the company said Wednesday.

Once the merger is complete, May Mobility said it will be the first public company in the U.S. that is focused entirely on autonomous ride-hailing vehicles. This is meant to differentiate it from a number of other public companies working on autonomy, including Tesla, Rivian, Alphabet (with Waymo), and trucking-focused Aurora and Kodiak.

The move sets up May Mobility to be a test of the stock market’s appetite for pure-play robotaxi ventures.

It will also be a test of May Mobility’s approach to autonomy, which it pitches as “asset-light” and “partnership-first.” Instead of owning and operating the robotaxis, May Mobility’s business revolves around selling its autonomous vehicles to its fleet partners over time while maintaining control of any remote supervision and software updates. In exchange, May Mobility receives either fixed fees or per-trip licensing fees.

Founded in 2017, May Mobility currently operates autonomous Toyota Siennas in three locations in the U.S. It has a partnership with Lyft in Atlanta, and offers rides in two cities — Eden Prairie and Grand Rapids — in Minnesota.

These deployments helped May Mobility generate around $10 million in revenue last year, with a cash burn of around $93 million. It has offered more than 550,000 paid autonomous rides to date, covering more than 1 million miles. The company recently spun up its first trial deployment in Japan and is planning commercial launches in Arlington, Texas, with Uber, at the end of this year or in early 2027.

May Mobility is merging with ACP Holdings Acquisition Corp., a special purpose acquisition company established by Houston, Texas-based investment management company Atlas Credit Partners. The merger will involve a $120 million “private investment in public equity” transaction, and up to $217 million from a trust account maintained by ACP Holdings — though shareholders in the SPAC may choose to redeem their stock at the time of the merger, which could reduce the amount of money going to May Mobility.

May Mobility said Wednesday that it will use the proceeds to fund more research and development, especially around removing its safety drivers, as well as supply-chain investments to reduce its bill-of-materials costs. The company is also targeting new geographic deployments, some of which it expects to announce later this year.

The world hasn’t been the same since ChatGPT made its debut in November 2022, when seemingly out of nowhere, a means of engaging with large language models propelled AI beyond hype and into day-to-day reality.

Robotics, like AI, has been discussed, built upon, and in some ways adopted for a long time now, but it’s yet to have that big breakthrough moment. And at TechCrunch Disrupt 2026, Nvidia Inception’s Global Head of Physical AI, Les Karpas, will dive into why that is and what could change that in his session on the Real World AI Stage, Robots Are Waiting for Their ChatGPT Moment. Here Is What Is Standing in the Way.

Image Credits:TechCrunch

After all, anyone who’s seen Nvidia CEO Jensen Huang’s keynotes the past few years knows the company is bullish about the robotics field. And you can get the opportunity to listen as Karpas details the limitations and opportunities in the AI and robotics spaces firsthand at Disrupt. From October 13-15 at San Francisco’s Moscone West, you and 10,000+ tech leaders, founders, and investors can catch Karpas and the rest of our extensive AI-focused lineup, which ranges from building companies to building in the real world.

Act now and get your ticket to save up to $200 before September 25, after which our prices increase. Save up to 30% when you bring a group, since the Disrupt experience is meant to be shared.

The all-too-real problem facing robotics

Karpas’ POV on the core limitation facing general-purpose robots is simple to understand but wildly complicated to resolve. Nobody has an internet-wide dataset for physical AI in the way OpenAI, Anthropic, and others had for language. Even self-driving fleets from Waymo and its competitors were able to build datasets through years of accumulated road miles, datasets that are still expanding market by market.

A growing ecosystem of startups is now trying to manufacture that scale artificially through simulation, synthetic data, and foundation models trained across many robot forms simultaneously. And that challenge of bridging the digital and physical is why we created the Real World AI Stage. Karpas will be joined by founders from Shield AI, Colossal Biosciences, FieldAI, and Foxglove, as they walk through the unique challenges and opportunities they face.

Nvidia’s unique perspective comes to Disrupt

Karpas is one of the best-positioned people in the world to answer the question his session presents. As Nvidia Inception’s Global Head of Physical AI, he coordinates the company’s relationship with the wide-ranging startup ecosystem stretching across robotics, automotive, manufacturing, mobility, smart cities, and much more. The same companies trying to solve that data gap are the ones he regularly interfaces with.

And his path to getting there is one that speaks to the cross-disciplinary skills the robotics industry demands. Across roles at Stanley Black & Decker, Intellectual Ventures, Herman Miller, iRobot, Cirque du Soleil, and more, he’s been an architect, manufacturing engineer, startup CEO, venture studio executive, and corporate venture capitalist.

Create your own breakthroughs at Disrupt

This is your chance to hear firsthand how Nvidia is thinking about the single biggest bottleneck in physical AI, and what it’ll take to clear it. That kind of insight could give you a leg up on your competition or spark an idea that could fuel your next startup venture. That kind of inspiration is what Disrupt is all about, and you can get it only by being in the room and making connections during the event.

Act swiftly to get the ticket that gives you the kind of access you need, whether it’s new sources of fundraising or new startups to invest in. Lock in your trip before prices increase on September 25. Bring a group and save up to 30% on bundle passes.

In Brief

Google says that a bug in its Pixel smartphones’ software was exploited in limited and targeted cyberattacks. The company said Tuesday that the bug, tracked as CVE-2026-58704, has now been patched.

According to the limited details about the vulnerability, the bug was found in Pixel phones’ modem, which lets the device connect to the internet. Exploiting the bug could allow an attacker to gain access beyond the sandboxed walls of the modem and into the broader phone’s data, a vulnerability known as privilege escalation.

The bug can be exploited silently and without any interaction from the phone owner in what’s known as a “zero-click” attack, meaning a victim does not need to click on a link or open a file.

Google did not say who was exploiting the bug, and a spokesperson for Google did not return a request for comment. It’s not uncommon for bugs like this one to be abused by surveillance vendors, such as spyware makers, who sell access to their data-stealing software to governments and law enforcement agencies.

In Brief

Amazon is raising the minimum starting pay for its U.S. core operations workers — like delivery drivers and fulfillment center employees — to $20 per hour and its average hourly pay to nearly $24 per hour. The company notes the higher pay represents an increase of $1 per hour for eligible employees.

“When you add in the value of our industry-leading benefits, average total compensation comes to more than $32 per hour,” Amazon Senior Vice President Udit Madan wrote in a blog post.

The investment in these raises comes out to over $1.5 billion, or about 0.06% of its $2.68 trillion market cap.

In addition, Amazon says it will now also provide employees with access to a low-cost banking system called Day 1 Financial, a membership in First Tech Federal Credit Union, as well as a 20% discount on purchases bought in-store at Whole Foods Market, the premium grocery chain that Amazon owns. If employees order groceries online via Amazon, they will also get a 10% discount.

Amazon already offers its employees free Prime memberships, prepaid education programs, and healthcare coverage, among its benefits.

On a subreddit for Amazon fulfillment center employees, people seem less than enthused. Regarding the Whole Foods discount, one worker commented, “So it would make it cost almost the same as a regular store, then? Hahaha,” a reference to Whole Foods’ historically higher prices.

Meta announced on Wednesday that its social network and X rival Threads is rolling out new tools that will allow podcast creators to promote their shows, grow their audiences, and engage existing listeners directly on the platform.

The news comes as X has also more recently been encouraging its creators to directly post videos on the platform, especially podcasts, to earn revenue and grow engagement. While Threads is not yet looking to host podcasts, it wants to make its platform a hub for posting and discussion.

To start, Threads will display a podcast profile card and banner on a creator’s profile and in their posts. Plus, when creators post episode links, they appear as tappable cards in the post, allowing users to easily listen.

Image Credits:Threads

The feature also lets podcast hosts directly post transcripts for listeners to read along with the episode, and they can tag featured guests, which displays a card related to that. Users can save any of these posts, making it easy to listen to the episode at a later time.

Threads will additionally be able to send reminders to creators when their new episode drops to post on Threads, which could help boost their engagement with listeners. The social platform is also providing tools that offer insights into how the audience is engaging with creators’ posts, with metrics like views over time by followers, non-followers, and all.

Image Credits:Threads

Last year, Threads rolled out new tools to increase discussion around podcasts, with features like better link cards and linking their profile with their podcasts. Threads is now close to X in terms of users, as Meta’s social network reached 500 million active users in June.

A great pitch doesn’t win Startup Battlefield 200. A great company does.

That’s why the VCs sitting in the judges’ seats matter so much. They aren’t looking for polished presentations or clever demos — they’re looking for founders solving real problems with businesses that can endure. Every question they ask is designed to uncover what slides and rehearsed answers can’t: conviction, clarity, execution, and whether this company has the potential to become the next category leader.

For more than 15 years, Startup Battlefield has helped launch companies that went on to define entire markets, and every year a new group of founders earns the chance to join that legacy.

Image Credits:Kimberly White / Getty Images

Today, we’re introducing the next wave of judges who’ll take their seats at TechCrunch Disrupt 2026, happening on October 13-15 at Moscone West in San Francisco. They’ll spend each day evaluating ambitious startups across industries, asking the tough questions founders hope for — and sometimes fear. If you want to understand how experienced investors really evaluate early-stage companies, there’s no better place to watch than Startup Battlefield, live on the Disrupt Stage.

Be sure to snag savings of up to $200 on your Disrupt 2026 ticket before rates increase on September 25 at 11:59 p.m. PT. Register here.

Meet the next five VCs judging Startup Battlefield 200

Now, without further ado, meet the next five judges joining this year’s Startup Battlefield, the ultimate startup pitch competition.

Image Credits:TechCrunch

Aditi Maliwal, General Partner, Upfront Ventures

Aditi Maliwal is a general partner at Upfront Ventures, where she invests at the pre-seed and seed stages in fintech, enterprise SaaS, AI applications, and developer tools. Her portfolio includes Clair, Writer, Arcade, and General Translation. Maliwal previously held roles at Google and Deutsche Bank and led the Series A investment in Chime during her tenure at a prior firm. She is Upfront’s first San Francisco-based partner.

Michael Palank, General Partner, MaC Ventures

Michael Palank is a general partner at MaC Ventures, a majority Black-owned venture firm that backs seed-stage consumer and enterprise companies through a culturally driven investment thesis, with a focus on AI, fintech, healthcare, media, entertainment, and aerospace and defense. Before transitioning to venture, Palank worked in Hollywood talent representation at William Morris and Overbrook Entertainment, Will Smith’s production company.

Nell Daly, Co-founder and Managing Partner, Revenge Capital

Nell Daly is the co-founder and managing partner of Revenge Capital, an evergreen sector-agnostic fund focused on backing overlooked founders — including women, BIPOC, LGBTQIA+, disabled, and neurodiverse entrepreneurs — across the United States and United Kingdom. Before launching the fund, Daly worked as a psychotherapist, a background she draws on directly in her founder-first investment approach.

Grace Ge, Partner, Amplify Partners

Grace Ge is a partner at Amplify Partners, where she invests in AI, developer tooling, data infrastructure, and modern SaaS — with a particular focus on the layer between foundation models and enterprise applications. Her portfolio includes Pinecone, Eppo, Vivun, Orb, and Avoca. Amplify’s broader portfolio includes Datadog, Fastly, Temporal, dbt, and Modal. Ge was named to the Forbes 30 Under 30 list for Venture Capital.

Crystal Huang, General Partner, Google Ventures

Crystal Huang is a general partner at GV, where she focuses on enterprise SaaS, AI infrastructure, and fintech — with a particular lens on product-led and developer-led go-to-market strategies. Her portfolio companies include Typeface, Pecan AI, Sardine, Cacheflow, PostHog, Roboflow, and Sift. Before joining GV, Huang was an investor at NEA and GGV Capital and worked in technology mergers and acquisitions at Blackstone.

Get to know all the VC judges on the Disrupt agenda.

Don’t miss Disrupt and Startup Battlefield 200

TechCrunch Disrupt 2026 is happening October 13-15 in San Francisco, bringing 10,000+ tech leaders, VCs, and founders together to meet the next generation of breakout startups, connect with the leaders who could change their startup’s trajectory, and get a front-row seat to where the industry is headed. Save up to $200 on your ticket before rates increase on September 25. Save even more when you register as a group.