Techin' Geek

AI agents are beginning to perform more advanced functions, such as locating suppliers that can transport goods, evaluating pricing options, and communicating with the supplier to arrange for delivery. However, they still require human intervention for payment processing related to shipments.

The modern financial landscape depends on financial rails, the foundational systems that facilitate the movement of money and information among banks, enterprises, and consumers. These financial rails were designed for human-driven transactions, rather than autonomous AI agents. For instance, conventional payment mechanisms like credit cards and ACH necessitate human approval for transactions, hindering the efficiency of agents crafted for autonomous operation.

A new startup, Natural, is addressing this challenge by overhauling the entire system from scratch. It has recently secured $30 million in new funding to embark on an ambitious strategy that positions it to directly compete with industry leaders such as Stripe.

Approximately a year ago, Natural’s co-founder and CEO Kahlil Lalji recognized that AI agents were advancing quicker than the current financial infrastructure, which fails to accommodate processes like autonomously paying a supplier, receiving payments, or conducting transactions amongst themselves.

Lalji possesses a background in banking and finance, but as he geared up to launch another startup, he aimed to steer clear of the financial sector. His previous venture, Ivella, a YC-backed financial product for couples, was sold in 2023 to Earnin, where he served as an engineer for two years. He conveyed to TechCrunch that he had faced setbacks in the finance domain following the conclusion of the Zero Interest Rate Policy era.

Nonetheless, Lalji could not overlook the opportunity.

“I kept on returning to it,” he stated. “It just seems evident that agentic payments are going to be fundamentally the most critical issue [in the] sector.”

Lalji joined forces with Eric Wang, his co-founder at Ivella, and Walt Leung, a former engineering manager at Nextdoor, to establish Natural in 2025. The startup positions itself as an agent orchestration platform that empowers AI agents to transfer and manage funds. By incorporating Natural’s framework, companies can enable their agents to perform autonomous payments, gather funds, and engage in transactions with both humans and other agents.

Natural captured the interest of Kirsten Green, founder and managing partner at venture capital firm Forerunner. Green, whose firm emphasizes consumer experiences and the future of commerce, led the company’s $30 million Series A funding round, increasing its total capital to $40 million.

Green was drawn to Natural’s expansive goals. The startup is not solely concentrated on assisting agents with purchasing and checking out goods for consumers; it also aims to transform payment infrastructure, including how disputes are managed.

While Natural has been in a beta phase until now, Lalji informed TechCrunch that the startup has made sufficient essential architectural choices to provide it with a “promising chance” to compete against established players like Stripe, which is also striving to reconfigure payment rails for AI agents.

Lalji anticipates that Natural’s rapid development pace will enable it to surpass established behemoths and construct the payment infrastructure that will function as the financial foundation for AI agents. The startup’s ambitions have attracted seasoned personnel who previously worked at fintech leaders Stripe, Ramp, and Square.

Although Natural considers Stripe its primary adversary, numerous other startups, including DCVC-backed Skyfire Systems, are attempting to reinvent the payments foundation for AI agents using USD-backed stablecoins. While Natural intends to integrate stablecoins into its system, it is also developing support for traditional banking transactions.

In the race to dominate this space, Lalji believes the entire market could experience substantial growth if transactions are conducted at computer speed instead of human speed. “The volume of payments that could happen globally might be two, three, or four orders of magnitude larger than the payments that occur today,” he remarked.

Paramount Skydance’s intended acquisition of Warner Bros. Discovery (WBD) has encountered an obstacle after a judge temporarily halted the transaction in light of a lawsuit brought by a coalition of 12 state attorneys general who contend that the merger would negatively impact competition.

U.S. District Judge Araceli Martínez-Olguín granted a 14-day halt on Monday following arguments from both parties last week. The coalition, spearheaded by California Attorney General Rob Bonta, might pursue another suspension after the initial 14 days, potentially further postponing the merger.

The states’ lawsuit claims that the merger would harm movie theaters, basic cable providers, and viewers. They assert that if the two entities merge, it would diminish competition in three key areas: wide release theatrical film distribution, “top-grossing” theatrical distribution, and basic cable licensing.

“This marks a significant first victory in our effort to prevent this megamerger from ever materializing,” said Attorney General Bonta in a statement. “History illustrates what occurs when a small number of individuals hold immense sway over markets crucial to the lives of Americans: reduced opportunities for the majority and inferior products and services for everyone. Through our lawsuit, we are advocating for a free and equitable market and a flourishing film and television sector that benefits both creators and audiences. We are well-prepared, have the law backing us, and are eager to continue presenting our case.”

The transaction would merge two prominent film studios along with streaming services Paramount+ and HBO Max. It would also result in one of the largest television network portfolios, uniting Paramount’s CBS and MTV with WBD’s CNN and HBO.

“We are confident the evidence will show that the State AGs’ antitrust claims lack merit, as their alleged markets and assertions of anticompetitive impacts are not grounded in current market realities,” a Paramount spokesperson stated to TechCrunch. “This merger is legitimate, promotes competition, and will benefit consumers, creators, workers, and the entertainment sector. We will continue to robustly defend the transaction and look forward to the hearings regarding the substance of the State AGs’ action.”

Paramount CEO David Ellison mentioned in May that the transaction was scheduled to finalize by September. The legal hurdle has the potential to hinder Paramount’s ambitions to become a significant competitor to companies such as Netflix.

The proposed acquisition has drawn scrutiny from filmmakers, actors, and industry professionals who contend that it would reduce competitive dynamics and further centralize the U.S. media industry.

WBD has not promptly responded to TechCrunch’s requests for comment.

Adobe is introducing fresh AI-driven functionalities to its beta iOS camera application known as Project Indigo, which debuted last year. The app originally provided professional controls, multi-frame super-resolution, and various capture modes, but is now enhancing its offerings with features that leverage LLMs (large language models) to evaluate photographs and suggest edits.

It is also incorporating additional AI capabilities, including sophisticated object removal, depth of field creation, and the option to apply varying styles to images.

Marc Levoy, the lead on Adobe’s project, had previously improved the photography technology for Pixel devices. He noted that many generative AI tools rely on prompt-based editing, and often, crafting the ideal prompt to alter a photo or achieve the desired outcome can be quite challenging. This is why many of the new experimental AI features consist of buttons designed to yield more consistent results.

For example, Google’s camera coaching feature for Pixel devices, which was released last year, provided more basic framing tips. In contrast, the functionalities of Project Indigo are quite detailed and could assist you in picking up some photography techniques, even if you do not fully agree with the AI’s evaluation.

Two features fall under this category. The first is the photo critique, which provides a “professional” assessment regarding composition, lighting, colors, and emotional resonance.

The second feature offers capture and edit suggestions, providing guidance on reshooting the photo by advising changes to composition, exposure, and objects in the frame. For example, the app recommended I eliminate a hexagonal white object that appeared in a photo I took.

Another section outlines how you can utilize an existing photo to enhance it using Adobe Lightroom controls.

Photography applications, such as Apple Photos, Google Photos, and Adobe Photoshop, have incorporated object removal capabilities for years. However, this feature generally requires users to circle or select an object by drawing on the display, which is not always foolproof.

Project Indigo’s newly introduced function offers toggles for elements to eliminate from an image, including background people, trash and bins, wires and poles, fences, vehicles, and other clutter. Users can also describe a custom object for removal. The outcomes of this feature are quite remarkable.

The app successfully removed my friend along with the item he was holding from the background without generating any odd artifacts.

The app also permits users to utilize AI for generating depth of field in a photograph, creating a simulated blurred background.

With this update, Adobe is also trialing a style transfer function that enables users to transform their images into various tones such as watercolor, pen and ink, ink line with color wash, monochromatic, and backlit subject. Some styles evoke memories of the early iterations of the Prisma app. With more advanced models, the outputs appear polished, though style transfer is not an innovative or particularly valuable addition.

Despite Levoy’s critique of the prompt-centric implementation, Project Indigo possesses a feature that allows users to articulate their edits. This can be employed to execute edit workflows that are absent in preset functionalities. However, this could lead to potential inconsistencies.

The company is deploying Google’s Gemini-based Nano Banana for these functionalities, while remaining open to integrating other models, including its proprietary Adobe Firefly model. These capabilities, gathered under the AI playground tab, remain in the trial phase, and only a limited user base will gain access to them. These features might never reach a broader audience, but it is encouraging to see enhancements that can educate users about the subtleties of photography, rather than merely generating more inconsistency.

Cybersecurity firms report that hackers are infiltrating websites running weak versions of the widely used blogging platform WordPress. An estimation suggests that the count of at-risk WordPress sites is in the tens of millions as of Monday.

Recently, WordPress addressed two major security vulnerabilities, urging website operators to upgrade their software “immediately.” The seriousness of these flaws led to WordPress implementing mandatory updates where feasible. Following this, cybersecurity agencies such as Patchstack, Hexastrike, and WatchTowr have cautioned that attackers are taking advantage of the vulnerabilities, actively compromising sites that have yet to update from compromised versions of WordPress. 

While the exact number of WordPress-based sites at risk remains uncertain, educated estimates can be made. The compromised versions of WordPress include 6.9.0 to 6.9.4 and 7.0.0 to 7.0.1. As per WordPress’ official data, there are upwards of 400 million sites operating on those flawed versions, though these figures likely don’t account for recently updated websites.

Cybersecurity expert Daniel Card, who analyzed a sample of about 3,500 WordPress sites, assesses that under 15% are at risk. If Card’s estimate is applied to the broader base of WordPress sites on the internet, the total could still reach around 90 million.

The researcher lauded WordPress for implementing auto-updates, Cloudflare for stopping attacks on at-risk sites, and sites leveraging cybersecurity measures like web firewalls for the minimal number of sites currently vulnerable to hacks. 

WordPress.org, the organization behind the development of WordPress’ open source code, did not swiftly reply to a request for comments. Megan Fox, spokesperson for Automattic, the firm responsible for WordPress.com and its contributions to the open-source initiative, informed TechCrunch that “all sites hosted by Automattic, including WordPress.com, Pressable, WPVIP, and WP.cloud partners, were secured prior to the release. When the code updates became available, we rolled them out immediately across millions of sites.”

One of the significant WordPress vulnerabilities was discovered and reported by Adam Kues of the cybersecurity company Searchlight Cyber, which named it WP2Shell. Along with the other flaw, hackers can gain complete remote access to at-risk websites.

A recent update indicates that YouTube is tightening its regulations regarding AI-generated content by specifying its monetization policies. To deter creators from relying on AI and other technologies for producing subpar content, YouTube’s revised policy outlines three classifications of videos categorized as “inauthentic content” that are ineligible for monetization.

YouTube previously had regulations concerning AI content. Last year, the platform revealed it was restricting creators’ capacity to earn revenue from what it terms “inauthentic content,” including videos that are mass-produced and other forms of repetitive content made simpler by AI technologies.

However, the latest change focuses on providing additional clarity to the existing guidelines.

The new guidelines categorize inauthentic content into three groups: generic, repetitive, or template-based material; followed by content that is distressing or off-putting; and finally, any content featuring AI personas discussing sensitive topics, such as health and finance.

While determining AI slop, like adult content, may feel subjective for many, YouTube’s regulations aim to clearly outline when the use of AI in content creation is permissible or not for videos monetized through its YouTube Partner Program (YPP).

The program, which allows creators to profit from ads and subscriptions, is essential to YouTube’s operations; permitting it to be overwhelmed with AI slop or other low-quality content would not align with the company’s financial interests. This is particularly significant, as YouTube competes closely with television and streaming services for advertising revenue. Google’s video platform currently generates more advertising income than its competitors and has surpassed Netflix in average global daily views.

As YouTube’s trust and safety head Matt Halprin mentioned in a Creator Insider video last week, the intent behind the policy revision is to reduce content farming practices — a nod to low-quality videos created solely for profit.

“AI can enable individuals to produce a plethora of videos,” Halprin noted. “Sometimes those videos are exceptional, significantly enhancing creativity. You can generate a greater volume of high-quality content that we want to promote and feature in YPP.”

“Nonetheless, that same new tool can also allow for the rapid creation of numerous very similar videos. They are generic, lacking a narrative arc, and fail to exhibit your creativity. Thus, the same technology can facilitate both remarkable content and that which aligns with content farming, and that’s the content we aim to exclude from YPP,” Halprin added.

The first category he mentioned includes repetitive content, such as material easily made with AI, computer-generated imagery (CGI), or templates, showing minimal variation across videos. This results in channels filled with generic videos. Halprin notes that tutorial videos could also fall under this policy if they replicate already existing content on the platform rather than being original.

The second category highlights what YouTube refers to as “off-putting” content, or material created to be distressing or emotionally manipulative to attract views. This might involve videos depicting an animal in distress, later rescued by someone, Halprin stated.

“We’ve received feedback from our viewers that they find this content undesirable. It doesn’t encourage them to return to that channel, or possibly even the platform,” Halprin remarked.

Channels focused on this type of content will be excluded from YPP, regardless of whether the content is AI-produced or not.

The third category directly addresses the use of AI personas — digital representations of real individuals made using AI. YouTube expresses its disinterest in incentivizing creators to employ AI personas for discussions about more sensitive topics, including finance, legal matters, healthcare, and medical issues.

Any YouTube channel exhibiting excessive amounts of any of these three content types will be ineligible for monetization, Halprin indicated. The policy clarifications were implemented on July 16 and affect all members of the YouTube Partner Program.

Nvidia’s leader, Jensen Huang, dedicated two days — July 15 and 16 — in Tokyo, engaging Japan’s elite in industry and chip supply, shortly after a keynote in Taiwan and several months post a visit to South Korea. He departed with agreements encompassing all of Japan’s technology landscape: a state AI factory, collaborations with top robotics firms in the country, and contracts with chip-material vendors fueling the next generation of Nvidia’s AI chips. His point was unmistakable. Nvidia is aiming at Japan’s manufacturing sector, and numerous major manufacturers in the nation are on board. The next phase of AI, Huang declared, revolves around factory settings, robots, and machinery, and he aims for Japan to spearhead this development.

Three decades ago, a $5 million investment from Sega kept Nvidia from collapsing; today, Nvidia and Japan’s industrial powerhouses require each other once more — this time to forge the physical-AI age, commencing with three initiatives:

Noetra — Japan’s sovereign-AI initiative. The nation prefers not to depend on American or Chinese AI for its factories and robotics. Thus, the government assembled about 44 domestic companies, including SoftBank, Sony, NEC, and Honda, to create its own AI tailored for robots, vehicles, and manufacturing environments. Tokyo is pledging up to 1 trillion yen ($6.2 billion) over five years, betting on “physical AI” developed locally, foundation models designed to operate machinery. Japan seeks control over the software intellect. However, the hardware necessary for its creation still comes from Nvidia. The U.S. chip giant is constructing “a Vera Rubin AI factory,” a vast data center filled with its advanced chips, anticipated to open in 2028, featuring 13,750 Vera CPUs and 27,500 Rubin GPUs, generating 140 megawatts. Noetra will manage this initiative, intending to develop the data center. The Noetra project unfolds in three phases: a reasoning model focused on Japanese-language proficiency starting in fiscal 2026; an omni-modal variant managing text, images, video, and audio by 2028; and “Real-world Native AI” designed for running robots by 2030, to be gradually released to external Noetra developers.

The robotics alliance — Japan’s industrial leaders rally around Cosmos. Nvidia is zeroing in on Japan’s manufacturing sector, and many of the leading robotics and production figures in the nation are signing up. Fanuc, Yaskawa, Kawasaki Heavy, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota, and the robotics consortium AIRoA have declared their intention to build upon Nvidia’s Cosmos models, an open-model initiative Nvidia launched in May with several international AI laboratories. In Tokyo, Nvidia provided them with a reason to commit, presenting Cosmos 3 Edge, a variation of the model that operates on its Jetson Thor chips integrated within the machines. Some are already experimenting with a shared control system; others, such as Honda R&D and Omron, are enhancing the tools presently. “The future of AI lies in the physical realm, and this is a once-in-a-lifetime opportunity for Japan,” Huang remarked in the company’s announcement. “Japan pioneered modern manufacturing. Now, it has a chance to revolutionize it for the era of intelligent industries.”

Toyota — vehicles and physical AI. Toyota incorporates Nvidia chips extensively throughout its systems. It dedicated its upcoming vehicles to Nvidia’s Drive platform at CES in January 2025; the updated work expands Nvidia’s reach into its manufacturing processes, where simulations are utilized to devise production lines, into the software governing its vehicles, and into systems interpreting road traffic. Toyota’s vehicles will utilize advanced driver assistance, which steers and brakes but still mandates a driver, a more cautious strategy compared to Waymo and Tesla, which are developing systems that rely less on human drivers.

Huang’s trip positioned physical AI at the core of Japan’s industrial strategy, and Tokyo is investing to support it. Confronted with a dwindling workforce, Japan aims for 10 million AI-equipped robots across 18 sectors by 2040, fueled by $65 billion in public and private investment in physical AI.

The long-term vision is even larger. Japan’s AI Robotics Strategy, unveiled in March, seeks to capture over 30% of the global AI robotics market by 2040, a market Tokyo estimates at around ¥20 trillion, or approximately $133 billion. METI is financing a domestic foundation model to operate the machines, with Noetra’s Nvidia-powered factory being the place where such expansive models, reaching into the trillions of parameters, would be trained. The belief is that Japan’s manufacturing data and base can catalyze advancement for physical AI.

Beyond industrial goals lies a sovereign aim. As the U.S. and China surge ahead in extensive AI, Tokyo desires its own data reservoirs, its own computing resources, and reduced reliance on infrastructure it cannot control. Huang appeared on July 16 alongside trade minister Ryosei Akazawa at the government’s physical-AI unveiling, with Prime Minister Sanae Takaichi participating via video link. The Takaichi administration has positioned AI and semiconductors at the heart of a growth strategy targeting ¥370 trillion ($2.3 trillion) in public and private investments by 2040. Noetra’s factory — marketed by Nvidia as “the world’s first national AI infrastructure” — signifies a notable commitment. Japan’s quest for self-sufficiency, for the time being, hinges on American chips.

In just two days, Huang engaged with nearly every pivotal figure in Japanese technology — the CEOs of Toyota, Fanuc, Yaskawa, Fujitsu, and Kawasaki during lunch, and numerous supply-chain leaders over skewers and whisky in a Kanda izakaya.

It mirrors the strategy he executed weeks prior — a homecoming keynote in Taiwan, fried chicken, and a 50,000-GPU agreement in Seoul last autumn. This time, it was Tokyo’s opportunity, encompassing robots, the supply chain, and the underlying chips.

Apple has recently initiated a lawsuit against OpenAI centered around trade secrets, alleging that the AI entity has engaged in a series of unethical activities to persuade both current and former Apple staff to divulge sensitive information. (OpenAI responded by stating they are “unaware of any evidence supporting the validity of this complaint.”)

In a recent edition of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I discussed whether this legal action might overshadow OpenAI’s highly publicized intentions to enter the hardware sector (with an initial focus on a mobile smart speaker) and pursue an IPO.

“Regardless of the possibility of the court granting injunctive relief or imposing any restraining order on OpenAI’s activities, it may naturally lead to a scenario that results in delays for OpenAI’s projects,” Sean remarked. “I imagine this was likely part of Apple’s rationale behind taking this step. They don’t make these moves haphazardly.”

Given the stakes involved, will OpenAI seek a swift settlement, or has it gleaned from its recent legal win against Elon Musk that it can weather the financial and reputational toll of a court trial? Kirsten, at the very least, anticipates the latter.

Continue reading for insights from our discussion, condensed for brevity and clarity.

Kirsten Korosec: Sean, what are your thoughts on Sam Altman potentially listening to you through a small device in your pocket?

Sean O’Kane: I’m fine with it. That may be predictable, but I’m good. No, thank you.

We’ll delve into it, no doubt, but this is supposedly the first product OpenAI is developing within its hardware division alongside Jony Ive and his team. They have been quite evasive since that peculiar video they released last year of them in a coffee shop or bar in San Francisco, speaking vaguely about hardware and traditional devices, such as laptops and phones. If this is indeed the path they are pursuing, all the best to those who wish to have someone like that perpetually eavesdropping. This isn’t for me.

Anthony Ha: It’s crucial to remember that with such devices, depending on their mobility, it isn’t just your voice being recorded, but also the voices of those around you. I might be okay with it — I’m not, but let’s assume I was — but if we met in person at Disrupt, it could be recording all of us. 

There will be numerous social norms that will need to be reevaluated if these devices become commonplace. I believe we should ridicule and criticize individuals who record others without their consent.

Kirsten: I’m bringing up this device which has been the subject of speculation for quite some time, and we’ll see what it truly turns out to be upon its official launch, but it holds significance in light of the lawsuit Apple filed last Friday. 

It was undoubtedly the most significant news of the week, and this is a lawsuit concerning trade secrets. It includes some rather astonishing claims, and we should highlight that these are allegations made in a complaint by Apple. The accusations against OpenAI involve a pattern of misconduct at the highest levels, specifically aimed at OpenAI personnel who were formerly employed at Apple. Notably, they’ve identified the chief hardware officer Tang Tan in this lawsuit.

This is crucial because Apple is charging OpenAI with essentially appropriating their trade secrets, potentially for use in a competing hardware product. I’m curious if we should skip over the merits of the lawsuit, given we haven’t completed full discovery, but what are your initial thoughts on the lawsuit aside from the fact that, wow, this is going to be entertaining?

Sean: Two points. First, this poses a significant risk to whatever OpenAI is currently developing. Even disregarding whether the court issues any form of injunctive relief or a restraining order regarding OpenAI’s actions, it inherently can result in situations causing delays in OpenAI’s projects, which I’m certain was part of Apple’s reasoning for this move. They are not akin to acting on a whim.

Secondly, we know that OpenAI has confidentially filed for an IPO. It might potentially happen as soon as the end of this year or early next year, if we take Sam Altman’s cautious rhetoric about the IPO into consideration. This raises numerous questions because, on one hand, we believe their current business is predominantly software; they aren’t really incorporating any hardware sector into that equation at this point.

As they prepare to approach the markets to pitch bankers and investors regarding their perceived addressable market, should they ascribe a substantial portion of that to a potential hardware division and products, this could significantly jeopardize that and alter the calculations regarding the IPO pricing. That’s where my thoughts stand.

Anthony: One [allegation] that I assume Apple must have significant data on is that they claimed more than 400 former Apple employees are now part of OpenAI. Admittedly, both organizations are very large, with thousands or even tens of thousands of employees. So, in percentage terms, it’s not an enormous figure. However, that does appear to be quite a large number and could indicate a serious talent exodus. 

Moreover, I’m curious, relating to Sean’s point, considering the potential IPO, how much marketing and brand damage did OpenAI ultimately incur from the previous trial it underwent? Although they seemed to have won, quite a bit of unsavory, albeit not catastrophic, information was revealed during the testimony. To what degree is their perspective now, “We want to avoid that experience again”? Or did they gain the insight that having encountered it once, they can endure it again and will be fine if faced with a trial against Apple?

Kirsten: I fully anticipate the latter, by the way.

Greetings again from TechCrunch Mobility, your central hub for the future of transport and the pivotal role AI plays now more than ever. To receive this in your inbox, enroll for free here — just click TechCrunch Mobility!

In my previous article, I discussed Uber and Waymo and how their collaboration seems to be unraveling. I forecasted that both companies would find themselves on conflicting sides when it comes to autonomous vehicle regulations. This wasn’t mere speculation.

Over the last few weeks, I’ve been conversing with sources and meticulously reviewing communications Uber sent to the D.C. Council, which is assessing a proposed legislation that would permit autonomous vehicles to operate within Washington, D.C.

My findings revealed: Uber and Waymo are already on divergent paths concerning the proposal, engaging in discussions both publicly and behind closed doors. Uber has presented a notably captivating argument to influence the framework governing autonomous vehicles.

Uber, which stands against the D.C. legislation, contends that it would replace human drivers for-hire and grant Waymo a de facto monopoly. Instead, it is advocating for a framework that mandates robotaxis to function on a ride-hailing network together with human drivers.

Insiders have indicated that the “hybrid” model has slim chances of becoming law. However, should it proceed, it would constrain AV innovators like Waymo to choose between two unfavorable options: either integrate their robotaxis within ride-hailing platforms like Uber or hire human drivers to accompany fleets of robotaxis that have taken years and massive investments to create.

A hearing at the D.C. Council on Monday saw representatives from Lyft, Tesla, Uber, and Waymo attending, alongside numerous disability rights and accessibility advocates, local business representatives, highway safety organizations, governmental officials, labor unions, and think tanks.

My insight — drawn from the public testimonies along with the calls and messages I received afterward — is that Waymo stands out as one of the few firms that generally supports the bill. Much of the rest of the industry is opposed.

Tesla’s senior policy advisor, India Herdman, echoed apprehensions I’ve heard from several AV developers, including objections to the 180-day, 250,000-mile mandatory testing criteria; the $1 million application charge; the $5 million permit fee; and the $0.15-per-mile tax. Tesla, among other firms, asserted that mileage accrued in other regions should contribute towards the mileage requirement.

Waymo, which has been trialing its AVs with human safety operators in Washington, D.C., has already exceeded the 180-day and 250,000-mile benchmarks. Therefore, if the bill were enacted as currently drafted, Waymo would have at least a six-month advantage entering the market.

Deals!

Uber is regarded as a giant in the realms of ride-hailing and delivery. It is currently solidifying that status through a $14.8 billion acquisition deal for Germany’s Delivery Hero. 

Should the acquisition finalize — and it will certainly require time to navigate through regulatory obstacles — Uber will have access to nearly 100 markets throughout Europe, the Middle East, Latin America, and Asia. The result: Uber’s delivery reach will be doubled.

Delivery Hero has also entered into a separate agreement to sell its business in 14 markets, where Uber Eats is already operational, to New York-based investment firm SSW Partners for $1.6 billion.

Other noteworthy deals include …

Self Inspection, a startup located in San Diego aiming to innovate the vehicle inspection process, secured $10 million in a funding round led by Sheryl Sandberg’s family office. Tire distributor U.S. AutoForce and automotive lender Westlake Financial made strategic investments. Early-stage funds Costanoa Ventures, Rebellion Ventures, and BrightCap Ventures also participated.

Senra, a startup that is modernizing the manufacturing of wire harnesses, raised $65 million in a Series B round co-led by Lowercarbon and Interlagos, with contributions from General Catalyst, Sequoia Capital, Andreessen Horowitz, and Founders Fund, among others.

Zepto, the Indian fast-delivery enterprise, is aiming for a valuation significantly below its previous $7 billion peak in its initial public offering, as reported by Bloomberg, referencing anonymous sources.

Notable reads and other tidbits

Chip Motors, a startup based in Miami, unveiled a low-speed compact EV tailored for short errands and families, incorporating some automated driving features.

The Los Angeles Police Department is reportedly terminating its partnership with Flock Safety, a surveillance firm that assists law enforcement in monitoring vehicles via thousands of license plate cameras distributed across the United States.

Lucid Motors is vigorously refuting a report that indicated the EV manufacturer was contemplating a Chapter 11 bankruptcy filing. The firm’s communications team, its CEO, and a filing with the U.S. Securities and Exchange Commission all assert that the rumors are unfounded. The original report caused the company’s stock to plummet over 50% on Tuesday, marking its largest intra-day decrease ever. The stock has since rebounded and is currently trading roughly 28% higher than it was before the significant drop. 

Lyft CEO David Risher claims it’s the “Good Uber,” according to Wired.

Manual, or traditional, transmission vehicles are fading away, as per preliminary government statistics indicating that only 0.6% of new vehicles manufactured for the U.S. in 2025 equipped with manual gear shifts, the Washington Post reported. I own two manual vehicles. Does that classify me as a driving unicorn?

The National Transportation Safety Board stated that the driver of a Tesla who crashed into a house in June had pushed the accelerator pedal to full throttle, bypassing the company’s Full Self-Driving (Supervised) software.

San Francisco mayor Daniel Lurie has encouraged state regulators to strengthen regulations on autonomous vehicles after Waymo robotaxis became immobilized in heavy July 4 traffic, ran out of battery, and obstructed key thoroughfares, exacerbating the congestion. In a letter (excerpts of which are included here) Lurie detailed four primary requirements he wishes to see implemented to guarantee robotaxi companies can “operate reliably” in extraordinary circumstances. 

SpaceX abruptly aborted its second attempt to launch the upgraded Starship rocket system on Thursday, just moments after the booster ignited at the company’s facility in South Texas.

Zoox issued a software recall following an incident in June where one of its robotaxis became perplexed by smoke emanating from an emergency fire scene.

One more thing …

Uber chief product officer Sachin Kansal discusses with TechCrunch EIC Connie Loizos about travel, AI agents, and straddling both sides of the robotaxi competition, in the latest Strictly VC podcast episode. If you prefer reading the dialogue, look at the Q&A.