Apple said that it reached out to OpenAI in February to share its concerns, but didn’t receive a response.

Apple, in a new lawsuit, has accused OpenAI of stealing trade secrets, as Big Tech races to develop next-generation consumer gadgets powered by AI.

In its suit filed last Friday (10 July), Apple levelled accusations of a “coordinated pattern of misconduct at an institutional level” against OpenAI, including at Io – the hardware start-up that the AI giant acquired last year.

Bloomberg reported that tensions between the two companies worsened after OpenAI enlisted Apple design visionary and Io co-founder Jony Ive to help with its product line. The company already owned around 23pc of Io prior to the acquisition.

Apple, according to the filing, said that it reached out to OpenAI in February to share its concerns regarding confidential information belonging to it, but didn’t receive a response. OpenAI, meanwhile, said that it has “no interest in other companies’ trade secrets”.

The lawsuit accused a number of former Apple employees who moved to OpenAI of a “pattern of theft” and alleged that OpenAI instructed potential job candidates from Apple to bring design artefacts and prototypes to their interviews. More than 400 former Apple workers are now employed at OpenAI.

Apple accused Chang Liu, its former senior system electrical engineer for iPhone, of leaving for OpenAI earlier this year without returning his devices or complying with exit procedures.

According to the filing, Liu allegedly told a former colleague that he still had access to an Apple-issued computer which he planned to access for sensitive information.

The lawsuit accused Liu of accessing his former colleague’s work computer without the company’s authorisation, and said that he exploited a previously unknown authentication bug without notifying Apple.

“He celebrated his improper access, exclaiming in a message left on his colleague’s Apple-issued work laptop, ‘LOL’, and that it was ‘so funny’”, the lawsuit read. It also accused Liu of accessing and downloading “dozens of Apple’s confidential hardware-related files” while developing hardware products for OpenAI.

The lawsuit further accused Liu of coaching a former Apple colleague on how to copy confidential information from Apple. “Knowing that OpenAI interviews would involve discussing Apple technology, Mr Liu advised her on which confidential Apple material about unannounced Apple products she should study before her interview,” the lawsuit read.

Another defendant, Tang Yew Tan, who currently works as OpenAI’s chief hardware officer, has been accused of “methodically using Apple’s confidential information to benefit OpenAI”.

Tan left Apple to co-found Io – alongside Jony Ive, Scott Cannon and Evans Hankey in 2024 – where he led the line as its chief hardware officer.

In his 20-plus-year stint at Apple, Tan also worked as the company’s VP of product design for iPhone and Apple Watches. Ive, Cannon and Hankey weren’t named in the lawsuit.

The lawsuit alleged that Tan, in the months before leaving for OpenAI, met with the company or its collaborators to discuss details of private meetings with an Apple supplier.

Tan allegedly used unauthorised access to seek information on an unannounced Apple product, the company said. He has also been accused of directing job candidates still working at Apple to bring confidential physical parts from products under development for ‘show and tell’ sessions at OpenAI.

“Other former Apple employees who had gone to work for OpenAI emailed themselves Apple’s confidential information to personal accounts on their way out the door,” Apple said.

“And others were improperly using their knowledge of Apple’s confidential and trade secret information to assist OpenAI in developing hardware.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

When humans mess with the sun in science fiction, it’s usually when a supervillain covers it up and imposes permanent darkness. A space tech company called Reflect Orbital wants to do the opposite: bring sunlight to the dark side of Earth with satellites equipped with giant mirrors. The FCC approved a single satellite as a test demonstration on Thursday, and some scientists are already unhappy about it. 

The approval green-lights Reflect Orbital to send its Eärendil-1 satellite into orbit. It’s a relatively small spacecraft, weighing 142 kilograms (313 pounds). 

Housed in its body is a thin-film square mirror measuring 18 meters by 18 meters (about 60 feet by 60 feet). The satellite is scheduled to launch into space on a SpaceX Falcon 9 later in 2026. 

Eärendil-1 promises to reflect sunlight onto Earth in a 3-mile circle that can be aimed basically anywhere that doesn’t have sunlight. The company has a web tool that shows you what this would look like, and it’s wide enough to light up entire neighborhoods, making it appear like daytime when it’s actually night. 

Ostensibly, this would be used to power solar panels at night, thus bypassing the one big drawback of solar panels: They can collect power only during the day. According to Reflect Orbital, electricity demand spikes right around sunset, which means power companies have to make the most power when the sun isn’t around to help. That increases fossil fuel usage, which is a contributing factor to climate change.

Reflect Orbital says that it wants to deploy 50,000 of these satellites in the event that the tech demonstration proves successful. That would put 16.2 million square meters of mirrors in Low Earth Orbit to light up large portions of the Earth on demand. For now, only the single Eärendil-1 satellite is approved for launch.

Scientists say this could be a disaster

Academics have been opposing the launch of Eärendil-1 since long before its FCC approval. Over 1,800 comments were made during the proposal stage, and most of them were negative. 

Researchers tend to agree that having 50,000 satellites beaming sunlight back to the Earth might be just as bad as a supervillain blocking out the sunlight entirely.

Multiple organizations filed complaints, including the American Astronomical Society and DarkSky, among others, citing potential problems with directing 16 million square miles’ worth of sunlight back onto Earth in the middle of the night.

“The concept of illuminating Earth from orbit represents a new category of artificial light at night with global ecological, cultural and regulatory consequences,” DarkSky said in an open letter to Reflect Orbital. “Based on the current scientific evidence, we do not see a viable pathway for this technology to align with responsible lighting principles or with our mission to protect natural darkness.”

Astronomers are also high on the list of people who oppose a big mirror brightening up the night, noting that even a single mirror flying in front of a telescope can easily overwhelm sensitive equipment in observatories on Earth. And since most astronomy research has to be done at specific times, thanks to the movement of the greater universe, missing an opportunity to observe something thanks to an army of mirrors may cause unrecoverable data loss. 

Scientists from other parts of the world are also unhappy with the notion that a single company in one country could ruin the sky for the rest of the world. 

“An individual RO satellite like Eärendil-1 is expected to have an optical brightness of at least 2 to 4 times that of the full moon,” the American Astronomical Society said in a complaint to the FCC. The AAS also notes that it’s not just the mirrors; because light scatters when it hits Earth’s atmosphere, the mirrors would also make the sky around them too bright to view, causing even more disruption in research. 

“For example, immediately adjacent to the sharp edge of the full moon, the sky is 10,000 times brighter than a dark sky with no moon,” the AAS said. “We expect a similar level of brightness surrounding Eärendil-1.”

Reflect Orbital admits that this is a regulatory blind spot, stating that there is “no established regulatory framework for space-based energy and lighting services.” The company says it’s open to regulation and working alongside scientists, while also stating that it intends to learn as much as it can from its test satellite before sending 50,000 more into space.

On Friday, Apple dropped the bombshell news it was suing OpenAI over the alleged theft of trade secrets, claiming that OpenAI stole Apple’s confidential data and engaged in efforts to learn proprietary information while recruiting former Apple employees.

In accusing OpenAI of stealing secrets about Apple’s unreleased products, Apple revealed that a former employee allegedly siphoned reams of sensitive files from the company’s shared network folders, weeks after leaving Apple for a job at OpenAI.

In its complaint, Apple says the former employee, a system electrical engineer named Chang Liu, allegedly “exploited a rare, previously unknown authentication bug” that allowed access to the company’s network. The bug is classified as a zero-day vulnerability, meaning that Apple had no time to fix it before it was allegedly exploited.

Apple has since fixed the bug and said it terminated the employee’s access once it learned of this “security breach.” In its complaint, Apple said the bug could have allowed a “few other” people to access data on its network, but alleged that only Liu exploited the bug to steal Apple’s confidential information while no longer an employee, citing a check of its server logs. 

The disclosure, while light in detail, highlights the challenges that organizations face with protecting sensitive corporate data after employees no longer work there. Companies often move to immediately cut off departing staff from further access to protect any sensitive information from leaving, including inadvertently. Companies that fail to fully decommission their employees’ accounts can face future security lapses, data breaches, or malicious actions by disgruntled staff.

Apple spokespeople did not respond to an email from TechCrunch with questions about the security vulnerability, how it was exploited, and when the company decommissioned the employee’s credentials.

“LOL… so funny.”

In the complaint, Apple alleged that Liu took “dozens of Apple’s confidential hardware-related files” over the course of several weeks while as a new OpenAI employee. 

Apple said the files contained “detailed information about unreleased products, engineering presentations, technical specifications, and proprietary project data.” 

The company claims Liu failed to return the Apple-issued work laptop he had previously used to access Apple’s network, suggesting it was once able to send and receive files from Apple’s internal systems. The complaint said that Liu allegedly claimed to have “another computer.” While he was at OpenAI, Liu also allegedly misused the access of an acquaintance, Yu-Ting Peng, a then-Apple employee who later went to work for OpenAI. Liu allegedly used Peng’s Apple-issued work laptop “while she was still employed at Apple and he was not.”

Apple said that during February 2026, Liu “tried to access Apple’s network storage — a cloud-based file repository containing Apple’s confidential engineering files, project documentation, and other proprietary information.”

Liu had allegedly discovered that he “still could access Apple’s network repository after leaving Apple, the result of a then-unknown authentication vulnerability.”

Apple did not describe the authentication “bug” that Liu allegedly used to access Apple’s network. However, authentication bugs generally refer to flaws in the login process that allow improper access to systems or data, either because of a weakness in how the login mechanism works or due to a misconfiguration, such as overbroad permissions or not decommissioning the login credentials of a former employee. 

Apple wrote in its complaint that when Liu learned he had unauthorized access to Apple’s systems, he did not report the bug to Apple under his employment agreement obligations, nor did he return his Apple-issued work laptop. 

The complaint added that Liu also failed to “delete the program that allowed the access” to Apple’s network. The company did not say what program or app that Liu allegedly used to access Apple’s systems. It’s not uncommon for employees to have tools, such as a work-approved VPN or remote-viewing app, that allow them to access sensitive data from outside of the company’s offices using their credentials.

Given that Liu was previously granted credentials to Apple’s network as an employee, TechCrunch asked Apple when the company decommissioned Liu’s access, but we did not hear back.

Once Liu allegedly gained access to the network share, he wrote to Peng: “LOL, I found out I can access the [network storage], so funny.”

Apple filed its suit in the U.S. District Court for the Northern District of California in San Jose, and has demanded a jury trial. OpenAI previously said it has “no interest in other companies’ trade secrets.”

The case, if it proceeds, could begin this year.

Meta’s acquisition of Manus drew widespread criticism in China for what was perceived as the handing over of critical technology to a geopolitical rival.

Chinese multinational technology company Tencent is in talks to become the largest external shareholder at AI start-up Manus in a deal that would further unwind Meta from a similar deal made previously, according to a report from Reuters, which cited two sources close to the matter. 

It was previously reported that US tech giant Meta would acquire Manus for roughly $2bn as the organisation aimed to further advance its work in the AI space. However, in April of this year, it was announced that Manus – which is based in Singapore, but originated in China – and Meta had begun dismantling the deal as a result of the Chinese government blocking it on the grounds of national security. 

When the deal was initially announced, it drew widespread criticism in China for what was perceived as the handing over of critical technology to a geopolitical rival.

An investigation commenced soon after, leading to the eventual dissolution of the agreement after the Chinese government blocked the deal.

Other investors reported to be involved in the new deal – which was first reported by the Financial Times – alongside Tencent are ZhenFund and HSG, both of whom were original investors in Manus. 

Investing in Manus is, for Tencent, an opportunity to further develop a reputation in the booming global artificial intelligence sector. In June, the organisation began testing a new AI assistant on its own app, WeChat, in an attempt to keep pace with its competitors in the space. 

WeChat users can interact with Xiaowei, Tencent’s AI agent, via text or voice prompts and the AI uses WeChat’s own large language model WeLM, while also tapping into DeepSeek to process additional queries.

Tencent has recently backed other innovators working in the AI space, such as DeepSeek, according to reports. In mid-June it was reported that Tencent was involved in a round in which the DeepSeek raised more than $7.4bn at a $50bn-plus post-money valuation. Tencent was also reported to have proposed taking a 20pc stake in DeepSeek. 

Updated, 2.43pm, 13 July 2026: This article was amended to clarify facts related to China’s investigation of the Meta-Manus acquisition, to clarify that ZhenFund and HSG were original investors in Manus before the Meta acquisition, and to clarify that Reuters and the Financial Times published the original reports.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Inconvo had offered configurable AI assistants connected to user-facing data, with the aim of allowing customers to analyse their data through a ChatGPT-like interface.

Attio, a London-based developer of AI-powered customer relationship management (CRM) software, has acquired Irish-founded AI analytics platform Inconvo for an undisclosed sum.

In a blogpost, Attio said that Inconvo’s founders, Eoghan and Liam Mulcahy, would be joining its engineering team in order to “expand Attio’s ecosystem of agents and integrations” and help to “make natural language interfaces ubiquitous in all of the tools” offered by Attio.

Inconvo’s independent platform has been shut down following the acquisition, according to a blogpost from Eoghan Mulcahy, who wrote that the founders’ goal in creating the platform two years ago was to get answers from data “by just asking, in plain language”.

He added: “Natural language access to data isn’t a nice add-on. It’s infrastructure, and it belongs right at the centre of the tools teams use every day.”

Inconvo had offered configurable AI assistants connected to user-facing data, with the aim of allowing customers to analyse their data through a ChatGPT-like interface.

“Eoghan and Liam are builders at heart and perfectly encapsulate the philosophy of our team and product,” said Nicolas Sharp, CEO and co-founder of Attio.

Inconvo was part of Y Combinator‘s summer 2023 batch and launched in 2024. Attio, founded in 2019, said that integrating Inconvo’s founding team would help it to build architecture that “lets teams move at a speed, scale and quality previously out of reach, with first-class agentic access wherever they operate”.

“Their vision to make natural language interfaces ubiquitous in all of the tools we use is one piece of the larger shift we’re driving at Attio – a CRM that transforms how revenue work gets done,” the UK company said.

Attio lists companies such as Modal, Snackpass and Union Square Ventures as users of its CRM platform.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Wilson is researching the role fathers could play in encouraging healthy mental development in adolescent boys.

Annie Wilson had a flourishing career in finance before she decided to pivot to psychology. Following the 2008 financial crisis, which marked a “significant turning point” in her life, Wilson returned to academia for her third degree – a bachelor’s in psychology. She previously held a bachelor’s degree in economics and master’s in business studies.

Wilson has committed fully to a career in psychology in recent years, and is currently working on her doctoral research into adolescent boys’ mental development with backing from the Craig Dobbin Doctoral Scholarship in Mental Health supported by the University College Dublin (UCD) Foundation.

What inspired you to become a researcher?

For me, research is the foundation upon which all credible psychological practice is built. Without it, clinical decisions would rest on intuition, tradition, or anecdote rather than evidence.

In child and adolescent psychology particularly, research has been transformative. It has established early interventions which in turn produces better long-term outcomes, identifying risk and protective factors for mental health difficulties.

I have always wanted to work with children and adolescents. The recent movement in the online world has made me think about how this will play out in society as children grow into adulthood.

Can you tell us about the research you’re currently working on?

I came interested in the online world, and how what children and adolescents are consuming online is changing how they fundamentally engage with their peers.

As I started to look into this area, I was drawn to boys specifically. The literature is starting to provide evidence around the stark disconnect between boys’ online world and their emotional wellbeing.

I became curious about what the landscape would like in 10 years’ time for adolescent boys now and how what they view now will change their fundamental behaviours and beliefs. It started to raise urgent questions about identity formation, mental health, and help-seeking behaviour.

Given boys already underutilise mental health services, and manosphere narratives that frame vulnerability as weakness, I wondered about long term interventions and could relational relationships move the needle?

After many conversations with my supervisors and my RSP (Research Studies Panel) panel, Gordon Harold, Brian O’Donohue and Marina Everri, it led me to think about where are the fathers in this equation. Could they be part of the solution?

This developed my research question for my scoping review which is, what is known in the existing literature about the relationship between father-son attachment and sons’ digital behaviour, the mediating effect of emotional regulation and what gaps exist in understanding this relationship as a basis for developing a targeted relational intervention?

In your opinion, why is your research important?

I believe it is fundamental to how relationships will develop in the coming years. We could be moving away from a more equitable experience for both men and women in western society.

As per the CyberSafeKids research conducted in 2024, 99pc of 12-14 have their own smart device. 38pc have experienced cyber bulling. 61pc have unrestricted access.

There are many facets to this, social comparison. Exposure to idealised images [has been] linked to body dissatisfaction and depressive symptoms. Teachers are concerned about the harmful/toxic content in student feeds.

We need to start to look at long term solutions to problematic social media use. As we wait to see how countries like Australia get on with the social media ban, we need to look for other solutions.

It will not reset with one intervention, we need a myriad of approaches to tackle the change that we are experience in adolescent boys content consume and their norms. I wonder does the relational relationship, hold the key to this need?

What commercial applications do you foresee for your research?

At present, I cannot see if there is a commercial application to the current research. However, that doesn’t mean it won’t have real-world value.

I can see it playing a meaningful role in shaping how we design digital literacy programmes, the kind that help young people, and particularly boys, develop a healthier relationship with technology.

Rather than a one-size-fits-all approach, the research points toward something more thoughtful – building skills gradually, layer by layer, in a way that actually sticks.

That kind of evidence-based framework could be genuinely useful to schools, parenting organisations, or anyone developing resources in this space.

What are some of the biggest challenges you face as a researcher in your field?

There are many challenges in this area, gaining a true reflection around how much adolescent boys behaviour has changed.

How understanding in real time, how the consumption of TikToks, reels, pornographic images and videos and how that will shape their future relationships and how they present in the world as an adult.

It is nuances and we are researching children and their parental attachment with their fathers. It will be difficult to gain quantitative research, I will be leaning more on qualitative findings through interviews and focus groups. We need to ensure we safe guard everyone who participates in the research so that they feel heard and understood.

Are there any common misconceptions about this area of research? How would you address them?

There are many common misconceptions in the father-son attachment research. The two main ones are that fathers play a peripheral role in the adolescent development. This is a myth.

Research consistently contradicts this, paternal involvement is independently associated with better mental health outcomes, stronger emotional regulation, and reduced risk-taking behaviour in adolescent boys, over and above maternal influence.

Another misconception, is that boys do not need emotional connection with their fathers.  Cultural narratives around masculinity suggest boys need discipline and challenge from fathers rather than emotional closeness.

In reality, adolescent boys with emotionally available fathers demonstrate greater psychological resilience, better peer relationships, and are significantly more likely to seek help when struggling.

We hope by looking at fathers and sons attachment, and investigating if modelling around rupture and repair can aid in emotion regulation, and could possibly safe guard boys from problematic social media use.

We want to develop an intervention that puts father and sons at the heart of the process. That we can develop a relational intervention, that is a sustained and a grounded framework for fathers to utilise.

What are some of the areas of research you’d like to see tackled in the years ahead?

Looking ahead, we’d love to see research that really reaches people, in the places where families actually live their lives: schools, community groups, and youth services.

One area we’re particularly passionate about is finding better ways to support boys and their fathers in navigating the digital world. Screen time and online culture are shaping how young men think, behave, and relate to others and we don’t yet have enough practical, real-world tools to address that.

We’d like to explore how father-son programmes could be woven into existing settings, whether that’s a school, a local sports club, or mental health services like CAMHS (Child and Adolescent Mental Health Services) and so that support is available to families who need it most, without requiring a whole new system to be built from scratch. That kind of joined-up, scalable approach feels especially important in a country like Ireland, where mental health resources are already stretched thin

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

The federal government is warning users of home and small office routers to secure their devices as Russia state hackers continue to mass-compromise them for use in obscuring nefarious actions against sensitive organizations in the public and private sectors.

Both the Russian and Chinese governments have been compromising routers for years, sometimes in prolonged tugs-of-war to wrest control of devices the other has already commandeered. The US government has occasionally issued covert commands and taken other steps to disinfect routers. Google and other companies have also worked to disrupt the massive botnets that control compromised routers in lockstep. The actions to date are little more than whack-a-mole exercises as the operators simply replace their botnets with new ones.

Proxy networks: The go-to tool

“Russian Federal Security Service (FSB) Center 16 cyber actors continue to exploit poorly configured and vulnerable networking devices worldwide, opportunistically compromising multiple critical infrastructure sector networks,” the Cybersecurity and Infrastructure Security Agency said Monday. The hacking groups are tracked under various names, including Berserk Bear, Energetic Bear, Crouching Yeti, Dragonfly, Ghost Blizzard, and Static Tundra. The advisory was co-issued by governments from around the world, including Australia, Denmark, New Zealand, and the UK.

The primary means of compromise the agency warned about was hackers scanning IP ranges with active Simple Network Management Protocol (SNMP) agents that accept common or default authentication credentials. These scans are run by the very sorts of router botnets the actors are trying to enroll the targeted device in. By sending malicious traffic from spoofed addresses, the hackers can use the SNMP agent on poorly configured routers to run malware. SNMP allows users to collect and organize information about managed networking devices or to modify that information to change device behavior.

systems

The last proprietary minicomputer, now in ‘deskside’ form if you fancy that

IBM has again teased small hardware, this time in the form of an update for its smallest POWER server.

The model S1112, teased Tuesday in a customer announcement, is a 2U, single-socket POWER11 server IBM offers in rack-mountable and what the company calls “Tower/deskside configuration.”

The rackable model can handle a ten-core POWER processor. The Tower/deskside form factor machine must make do with a four-core engine

IBM seems to have two roles in mind for the new machines: edge deployments and standalone use by those who are taking their first strides into using the last remaining proprietary minicomputer ecosystem.

One is edge deployments. The other is as an entry-level box, with the description of the tower unit suggesting its very existence means “even the smallest customers” can use it as an on-ramp to more POWER implementations.

News of the S1112 marks the second time in a week that IBM has gone low with modest hardware. Last week Big Blue teased the z17 ME2, a rackable mainframe that it said completed its range by offering a smaller and cheaper piece of hardware.

The twin launches continue IBM’s policy of creating smaller versions of its enterprise hardware, albeit well after launch of big iron: the first POWER 11 boxes landed in July 2025 and the first z17 series mainframes debuted in April of the same year.

The S1112 includes a quartet of DIMM slots and can handle up to 512GB of DDR5 memory. The box runs IBM i, AIX, and Linux – or all three because it supports IBM’s PowerVM virtualization tools.

In an almost certain non-coincidence, Big Blue on Tuesday also announced upgrades for PowerVM including improved automation and support for the S1112.

Big Blue has also looked after users of bigger POWER fleets, by expanding the number of Spyre accelerators – IBM’s neural processing units – that POWER servers can support from eight to twelve. IBM pitches POWER as a capable AI platform, so allowing it to use more accelerators can’t hurt.

IBM plans to start selling most of the kit described above on July 24, although customers who crave the S1112 to deploy in Taiwan will have to wait until September. 15. Would-be buyers in South Africa, India, and China must wait longer still, until December 11. ®

Tesla has said Cybercab employee rides are starting soon at its Texas factory. The announcement came with a clip of a gold Cybercab, butterfly doors up and no steering wheel or pedals, driving itself across the outbound lot, Mashable reports.

Note the tense. Tesla said the rides are “starting soon”, not that they have started, and several outlets have already reported it the other way round.

That is the whole announcement. No route, no fleet size, and no word on whether the rides happen on public roads or entirely on Tesla’s own property.

What we are actually looking at

Strip away the framing and a car drove itself around a car park. Autonomous vehicles have managed that on private land for well over a decade.

Giga Texas is an enormous site with real internal roads, so a genuine campus shuttle would mean something. A loop across the outbound lot would not.

Tesla has not said which it is. Electrek, which broke down the announcement, concluded it could not tell either.

What it is definitively not is the thing people are waiting for. The Cybercab has not joined Tesla’s paying robotaxi fleet in Austin, which still runs on Model Y vehicles.

The hardware is not the problem

It is worth being fair to Tesla here, because the manufacturing is genuinely impressive. Its unboxed assembly process works, and the Cybercab is the most efficient vehicle the company has built.

More than a hundred finished Cybercabs are reportedly stacked in the Giga Texas lot. Tesla can make these things.

The bottleneck is software, and it is a brutal one. With no steering wheel and no pedals, there is no fallback, so if the self-driving system fails the vehicle simply cannot be driven.

That is survivable at walking pace in a car park. It is a different proposition on a public street.

The record so far

Tesla’s autonomy numbers remain difficult. TNW has reported that its Austin robotaxis crash roughly every 57,000 miles, about four times worse than the human average.

The scale gap is starker still. Tesla has 42 vehicles authorised for driverless ridehailing in Texas against Waymo’s 577, which is a matter of public record rather than opinion.

Progress is real, though, and worth saying plainly. Tesla began on-road engineering tests of the production Cybercab in Austin at the end of June, with a safety monitor riding in the passenger seat, and it now runs robotaxis in Miami with no safety monitor at all.

The story that matters

Here is what got buried under the video. On 9 July, two days before the clip appeared, the NHTSA administrator said the agency would “absolutely” consider scrapping the rule requiring driverless vehicles to have steering wheels and manual controls.

That is the constraint that has kept the Cybercab in a legal grey zone. A car with no controls is difficult to square with federal standards written on the assumption that a human might need to take over.

The dismantling is already under way. Washington has proposed dropping the brake pedal requirement for autonomous vehicles, and is now weighing the steering wheel rule too.

Tesla’s strategy suddenly makes sense in that light. It has pointedly refused to apply for individual exemptions, which are capped at 2,500 vehicles a year, betting instead that the rules themselves would move.

So far, that bet is paying off. Musk has been explicit that broad regulatory approval, not exemptions, is how the Cybercab reaches scale.

Why the car park video exists

A company mass-producing a vehicle it cannot yet legally or reliably deploy needs to show motion. A gold Cybercab gliding past the factory doing something, anything, is motion.

It is not dishonest, exactly. It is just a much smaller milestone than the reaction suggests, and Tesla left the framing conveniently vague.

The serious test is unchanged. Get a Cybercab carrying paying passengers on public roads in Austin, with nobody in the front seat.

Until then, the most consequential thing happening to this vehicle is not being filmed in a car park. It is being written into the Federal Register.