Techin' Geek

Cybersecurity experts have revealed a series of cyber assaults aimed at Apple users globally. The methods employed in these hacking operations have been named Coruna and DarkSword, with both government operatives and cybercriminals utilizing them to extract information from individuals’ iPhones and iPads. 

It is uncommon to witness extensive hacks targeting users of iPhones and iPads. Over the past ten years, similar incidents have primarily involved attacks against Uyghur Muslims in China and individuals in Hong Kong.

Now, portions of these robust hacking instruments have surfaced online, potentially jeopardizing hundreds of millions of iPhones and iPads operating outdated software to data breaches.

We are dissecting the available information about these recent threats to iPhone and iPad security, as well as the protective measures you can undertake.

What are Coruna and DarkSword?

Coruna and DarkSword are two collections of sophisticated hacking toolkits that encompass various exploits capable of infiltrating iPhones and iPads to extract sensitive data, including messages, browsing history, geolocation, and cryptocurrency information.

The cybersecurity professionals who identified these toolkits report that Coruna’s exploits can compromise iPhones and iPads operating on iOS 13 through iOS 17.2.1, launched in December 2023. 

Conversely, DarkSword includes exploits that can breach iPhones and iPads with the latest models using iOS 18.4 and 18.7, released in September 2025, according to Google cybersecurity analysts examining the code.

However, the danger posed by DarkSword is more pressing for the general populace. A portion of DarkSword has been leaked and uploaded to the code-sharing platform GitHub, allowing anyone to access the harmful code and initiate attacks against Apple users on older iOS versions. 

How do Coruna and DarkSword work?

These types of attacks are inherently indiscriminate and perilous, as they can ensnare any individual who visits a specific website hosting the harmful code.

In some instances, victims can be compromised merely by accessing a legitimate website under the management of malicious hackers.

Upon initial infection, Coruna and DarkSword capitalize on several vulnerabilities within iOS, enabling hackers to gain near-total control over the target device, consequently allowing them to extract the user’s private information. This data is then transmitted to a web server managed by the hackers. 

At least some components of the Coruna toolkit, as previously reported by TechCrunch, were initially created by Trenchant, a hacking and spyware division within the U.S. defense contractor L3Harris, which sells exploits to the U.S. government and its leading allies.

Kaspersky has also associated two exploits in Coruna’s toolkit with Operation Triangulation, a complex and likely government-led cyber operation allegedly conducted against Russian iPhone users.

After Trenchant created Coruna — the specifics remain unclear — these exploits seemingly reached Russian spies and Chinese cybercriminals, potentially through one or several intermediaries selling exploits on the dark web. 

Coruna’s trajectory exemplifies once again that formidable hacking tools, even those generated for the U.S. under strict confidentiality protocols, can leak and spread uncontrollably. 

An instance of this occurred in 2017 when an exploit devised by the U.S. National Security Agency, capable of remotely infiltrating Windows computers globally, leaked online. The same exploit was later utilized in the destructive WannaCry ransomware assault, which indiscriminately hacked hundreds of thousands of computers worldwide. 

In DarkSword’s case, researchers have tracked attacks targeting users in China, Malaysia, Turkey, Saudi Arabia, and Ukraine. It remains uncertain who originally crafted DarkSword, its transfer to various hacking factions, or the manner of its leak online.

It is unclear who disseminated and uploaded the tools to GitHub, or their motives.

The hacking instruments, which TechCrunch has reviewed, are composed in web languages HTML and JavaScript, allowing them to be relatively simple to configure and self-host by anyone wishing to conduct malicious attacks. (TechCrunch is refraining from linking to GitHub as the tools can be employed in harmful attacks.) Researchers on X have already attempted the leaked tools by hacking their own Apple devices with vulnerable versions of the company’s software.

DarkSword is now described as “essentially plug-and-play,” as articulated by Justin Albrecht, principal researcher at the mobile security firm Lookout, to TechCrunch. 

GitHub informed TechCrunch that it has not removed the leaked code but will retain it for security research purposes.

“GitHub’s Acceptable Use Policies prohibit posting content that directly supports unlawful active attacks or malware campaigns causing technical harm,” Jesse Geraci, GitHub’s online safety counsel, informed TechCrunch. “However, we do not prohibit the posting of source code that could be used to develop malware or exploits, as the dissemination of such source code has educational merit and ultimately benefits the security community.”

Is my iPhone or iPad vulnerable to DarkSword?

If your iPhone or iPad is outdated, you should strongly consider updating it without delay.

Apple has advised TechCrunch that users operating on the latest versions of iOS 15 through iOS 26 are already shielded.

According to iVerify: “We highly recommend updating to iOS 18.7.6 or iOS 26.3.1. This will mitigate all vulnerabilities exploited in these attack vectors.”

Apple’s own data indicates that nearly one in three iPhone and iPad users are still not utilizing the latest iOS 26 software. This suggests that potentially hundreds of millions of devices remain susceptible to these hacking instruments, as Apple estimates over 2.5 billion active devices globally. 

What if I can’t or don’t want to upgrade to iOS 26?

Apple also mentioned that devices using Lockdown Mode, an optional enhanced security feature first implemented in iOS 16, can also obstruct these particular attacks. 

Lockdown Mode is advantageous for journalists, dissidents, human rights advocates, and anyone who believes they may be targeted based on their identity or occupation. 

Although Lockdown Mode is not flawless, there is no public evidence suggesting that hackers have been able to bypass its protective measures thus far. (We inquired with Apple about whether that assertion still holds true, and will update if we receive a response.) Lockdown Mode has reportedly thwarted at least one attempt to install spyware on a human rights defender’s device.

The surveillance technology sector is currently under scrutiny, and not for commendable reasons. Amid controversy regarding the U.S. Immigration and Customs Enforcement’s use of Flock’s camera network for monitoring individuals, as well as home security camera manufacturer Ring facing backlash for developing new functionalities that would allow law enforcement to request footage from homeowners about their surroundings, there is an extensive discussion surrounding safety, privacy, and the dynamics of surveillance.

However, controversy does not negate market opportunities, and the ongoing advancements in vision-language models have further propelled companies that create novel ways to assist organizations in overseeing activities within their facilities.

Matan Goldner, co-founder and CEO of the video surveillance startup Conntour, emphasizes the significance of ethics in this domain, claiming that his company carefully chooses its clientele. While this might not be perceived as typical startup logic for a company that is still in its early stages, Goldner asserts that Conntour is in a position to do so as it already boasts a number of substantial government and publicly traded clients, including Singapore’s Central Narcotics Bureau.

“Our relationship with such prominent clients enables us to be selective and maintain control […] We truly have autonomy over who utilizes our system, the intended applications, and we can determine what we deem to be ethical and lawful. We apply our discretion and assess specific clients that we’re comfortable collaborating with, based on our understanding of their intended use,” Goldner shared with TechCrunch in an exclusive discussion.

This momentum has assisted Conntour in more than just selecting clients. Investors have taken notice: The startup recently secured a $7 million seed funding round from General Catalyst, Y Combinator, SV Angel, and Liquid 2 Ventures.

Goldner noted that the funding round was concluded within a mere 72 hours. “I think I organized around 90 meetings in about eight days, and just after three days — we started on Monday and wrapped up by Wednesday afternoon,” he explained.

Nevertheless, Conntour’s cautious approach seems wise, particularly considering the immense power of AI tools within this sector. The company’s video platform employs AI models to allow security personnel to query camera feeds using natural language to detect any item, individual, or scenario in the footage, in real time — akin to a Google-like search engine tailored for security video feeds. It can autonomously monitor and identify threats based on predefined rules while generating alerts automatically.

In contrast to traditional systems that rely on predefined definitions or parameters to identify specific objects, movement patterns, or behaviors, Conntour asserts that its system leverages natural and vision language models, giving it a significant level of adaptability and user-friendliness. A user can simply request, “Identify instances of someone in sneakers handing over a bag in the lobby,” and Conntour’s system will swiftly comb through all recorded footage or live video feeds to provide pertinent results.

Furthermore, thanks to the integration of AI models, users can effortlessly pose questions regarding the footage and receive textual answers, along with the corresponding video feeds, and even create incident reports.

The key advantage of the company, however, lies in its scalability. Goldner clarified that the platform distinguishes itself from other AI video search solutions by being engineered to efficiently accommodate systems made up of thousands of camera feeds. Notably, he mentioned that Conntour’s system can oversee up to 50 camera feeds using a single consumer GPU, such as Nvidia’s RTX 4090.

The company achieves this by employing a variety of models and logical systems, and then determining which models and systems the algorithm should apply for each query to minimize computational demands while delivering optimal results.

Conntour asserts that its system can be entirely implemented on-site, completely in the cloud, or a combination of both. It can integrate with most existing security infrastructures or function independently as a comprehensive surveillance platform.

Nonetheless, a persistent issue in the video surveillance industry remains: The effectiveness of surveillance is only as good as the quality of the captured footage. For instance, distinguishing details from low-resolution footage taken in a dimly lit parking area with a dirty lens is quite challenging.

Goldner states that Conntour mitigates this challenge by providing a confidence score alongside its search results. If a camera feed’s source is of inadequate quality, the system presents results with low confidence levels.

Looking ahead, Goldner identifies the primary technical challenge as integrating the full extent of LLM capability within the system while retaining efficiency.

“We are focusing on two objectives simultaneously, which tend to contradict each other. On one side, we aspire to offer complete natural language flexibility, LLM-style, allowing for any inquiries. Conversely, we are concerned with efficiency, aiming to utilize minimal resources, as processing [thousands] of feeds is extremely demanding. This conflict represents the most significant technical obstacle we face in our field, and it’s what we are diligently striving to resolve.”

On Thursday, the enterprise AI firm Cohere unveiled its inaugural voice model: Transcribe, an open-source automatic speech recognition model suitable for applications such as note-taking and speech evaluation.

Weighing in at only 2 billion parameters, the model is designed for use with consumer-grade GPUs for those who prefer self-hosting. It presently accommodates 14 languages: English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Chinese, Japanese, Korean, Vietnamese, and Arabic.

According to Cohere, Transcribe outperforms models like Zoom Scribe v1, IBM Granite 4.0 1B, ElevenLabs Scribe v2, and Qwen3-ASR-1.7B Speech in the Hugging Face Open ASR leaderboard, registering an average word error rate (WER) of 5.42, the lowest among all models tested.

The company asserts that Transcribe achieved an average win rate of 61% against competing models when human judges evaluated its transcriptions based on accuracy, coherence, and usability. Nevertheless, the model lagged behind its competitors in transcribing Portuguese, German, and Spanish.

Cohere notes that Transcribe can handle 525 minutes of audio in just one minute, which is impressive for its model category.

The firm intends to incorporate Transcribe into its enterprise agent orchestration platform, North, and is offering the model for free via its API. Additionally, it will be accessible on Model Vault, Cohere’s managed inference platform.

The popularity of speech recognition models is on the rise as the need for note-taking and dictation applications like Granola and Wispr Flow increases.

Earlier this year, Cohere reportedly informed investors that it was on track for an annual recurring revenue of $240 million by 2025, with CEO Aidan Gomez stating that the startup might go public “soon”.

WhatsApp is introducing a range of new functionalities and updates, including one that creates AI-driven suggested responses based on your chats. The messaging service, owned by Meta, is also debuting additional features, such as a new method for clearing storage, capabilities for enhancing photos with Meta AI, and more.

Most significantly, however, is the application’s enhancement to its “Writing Assistance” function that aids users in composing messages. Writing Assistance, which was first introduced last August, already helps users reinterpret, proofread, or modify the tone of their communications. On Thursday, the Meta-owned firm announced in a blog post that this recent enhancement will assist users in perfecting their messages.

Meta likely anticipates that users will leverage its in-app features for message preparation instead of external services like ChatGPT. Naturally, not everyone may be inclined to utilize the new feature, as users probably favor genuine, personal exchanges with friends and relatives over AI-generated texts. (Employing AI for drafting an email is one thing; deploying it for family group chats is entirely different!)

The company asserts that conversations remain confidential, even if individuals utilize Writing Assistance.

To access this feature, users must tap the chat box, select the stickers icon within the typing area, and then click on the pencil with a sparkledust icon indicating an AI functionality.

Regarding storage management, WhatsApp can now assist users in locating and eliminating large files directly within any conversation. This allows users to remove unnecessary items without erasing entire discussions. For example, they can opt to only delete media files when clearing a chat while preserving the chat history.

The company also revealed that users can now employ Meta AI to enhance photos right within a chat. This enables users to perform actions like removing distracting elements from an image, altering the background, or applying a new aesthetic.

Moreover, WhatsApp now facilitates the transfer of chat history from iOS to Android, as well as within the same operating system. Additionally, users now have the capability to be logged into two WhatsApp accounts simultaneously on iOS, a feature already available on Android.

Stickers on WhatsApp are also receiving an update, as WhatsApp will now recommend specific stickers to users while they type emojis, allowing them to replace them with a sticker.

These new functionalities are currently being rolled out and will soon be accessible to all users, according to WhatsApp.

Recent research from Anthropic indicates that although AI is swiftly altering the landscape of work, it has not substantially reduced the number of jobs — at least, not at this stage. However, according to Peter McCrory, Anthropic’s head of economics, while the labor market remains “still healthy,” early indications show varying effects, particularly for younger individuals just starting their careers. 

During a discussion at the Axios AI Summit in Washington, D.C., McCrory mentioned that the company’s latest economic impact analysis reveals minimal indications of large-scale job displacement thus far. 

“There is no significant difference in unemployment rates” between individuals using Claude for the “most essential functions of their jobs in automated manners” — such as technical writers, data entry personnel, and software developers — and those in less AI-exposed occupations requiring “physical interaction and dexterity in the real world.” 

However, given the rapid integration of AI across various sectors, this situation could change rapidly. If Dario Amodei, CEO of Anthropic, is accurate, AI could potentially eliminate half of all entry-level white-collar positions and escalate unemployment to as high as 20% within the next five years.

“Displacement effects could emerge very rapidly, so it’s essential to implement a monitoring framework to understand that before it occurs, enabling us to catch it in real-time and ideally pinpoint the right policy response,” McCrory shared with TechCrunch.

Understanding those trends is crucial, he stated, highlighting the importance of monitoring AI’s growth, adoption, and diffusion.

In theory, McCrory noted, AI models like Claude are capable of performing nearly any task a computer can manage. However, in reality, most users are merely beginning to tap into those functionalities.

He mentioned that Anthropic examined which job roles involve tasks that AI excels in, are currently being automated, and are associated with actual workplace applications — these are the areas most likely indicative of where displacement could arise. 

Anthropic’s fifth economic impact report, released on Tuesday, also indicated that even in sectors where minimal displacement has occurred, a widening skills gap is emerging between early Claude adopters and newcomers.

Those who adopted the technology earlier are more likely to derive significantly greater value from the model, utilizing it for work-related tasks instead of casual or sporadic uses, and applying it in more advanced ways, such as acting as a “thought partner” for iteration and feedback. 

McCrory remarked that these results imply AI is evolving into a technology that benefits those already skilled in its use — and that individuals who can successfully incorporate it into their workflows will increasingly maintain a competitive advantage.

This benefit is also not uniformly distributed geographically. The report further revealed that “Claude is utilized more intensively in high-income nations, within the U.S. in areas with a larger concentration of knowledge workers, and for a relatively narrow range of specialized tasks and professions.”

In essence, despite claims of AI serving as an equalizer, its adoption may already be skewing in favor of the affluent, and could further amplify those advantages as proficient users advance more rapidly.

The creator of the spyware company Intellexa has announced intentions to contest a conviction issued by a Greek court related to accusations that he and three other executives unlawfully acquired personal information during a large-scale wiretapping operation in the nation.

The surveillance controversy, often dubbed “Greek Watergate,” involved the infiltration of numerous devices owned by high-ranking Greek government officials, opposition leaders, military personnel, and reporters through the use of Intellexa’s Predator spyware. This software can access both iPhones and Android smartphones to extract call history, text conversations, emails, and location details, typically by deceiving a target into clicking a harmful link.

Numerous high-ranking officials within the Greek government, including the director of Greece’s national intelligence agency and a key advisor to Prime Minister Kyriakos Mitsotakis, stepped down following disclosures that several journalists’ phones had been monitored. Thus far, no government officials have faced conviction regarding the surveillance, and dissenters have alleged that the Mitsotakis administration is engaging in a cover-up.

The founder of Intellexa, Tal Dilian, received a conviction in February and was sentenced to eight years behind bars. In a statement initially reported by Reuters on Wednesday, he expressed that he would not serve as a “scapegoat.”

Regardless of whether Dilian is, as he asserts, a scapegoat, the comment stands as the clearest indication yet from anyone associated with Intellexa that the Mitsotakis administration sanctioned the hacks.

“I contend that a conviction without proof is not ⁠justice; it may be part of a cover-up and potentially a criminal act,” Dilian articulated to Reuters. He mentioned his readiness to provide evidence to both national and international regulatory bodies.

Dilian did not reply to TechCrunch’s inquiry about his comments. The Greek embassy in Washington, D.C. has yet to respond when reached for comment by TechCrunch.

Dilian also informed Reuters that surveillance tools like Predator are generally sold strictly to governments, which bear responsibility for their lawful deployment.

In 2024, the U.S. government placed sanctions on Dilian after it was discovered that Predator was used against the phones of U.S. officials and journalists. These sanctions effectively prohibit any business dealings with Dilian and his other sanctioned commercial partners.

In August of last year, a blaze swept across 10 acres of grassland flanking California’s I-280 close to Redwood City. As firefighters battled the fire, traffic came to a standstill, and officers from the California Highway Patrol (CHP) instructed drivers to turn around and take the freeway exit in the opposite direction. 

Among those drivers was a new hurdle: a Waymo robotaxi.  

Video of the event captures the Waymo AV attempting to navigate around stalled vehicles by driving on the shoulder, ultimately having to reverse from the approaching wrong-way traffic before coming to a complete stop. 

Despite intervention from the company’s remote assistance team, the robotaxi remained stationary. Consequently, Waymo resorted to a solution that has proven effective before and dialed 911. 

“Highway patrol turned everyone around, but regrettably our vehicle can’t turn around,” one of Waymo’s remote assistance agents told a local 911 dispatcher, as per a recording acquired by TechCrunch through a public records request. The agent requested that officers onsite transport the robotaxi away and arrange for the passenger’s transport. 

Approximately 30 minutes post-call to 911, a CHP officer entered the vehicle and transported the robotaxi to a park-and-ride lot adjacent to the highway, according to a CHP incident report obtained by TechCrunch. Subsequently, one of Waymo’s “roadside assistance” personnel drove it from that location, the company conveyed to TechCrunch. 

The Redwood City occurrence could be regarded as an atypical event, a somewhat embarrassing yet unavoidable hiccup in Waymo’s swiftly growing robotaxi service network.

However, this was not a standalone incident. Waymo has depended on taxpayer-supported first responders to maneuver its vehicles when complications arise, even with its own roadside assistance team in place. In at least six cases identified by TechCrunch, first responders were required to take control of Waymo vehicles and relocate them from traffic during emergencies, including one situation where an officer was engaged in responding to a mass shooting.

Recently, Waymo has faced backlash from lawmakers regarding its employment of remote assistance staff, including some who work from the Philippines, to assist its robotaxis in determining the most suitable route through complicated circumstances. The roadside assistance team has garnered far less attention.  

Representatives from the company did not mention the roadside assistance team during a contentious March 2 hearing in San Francisco concerning the behavior of Waymo’s robotaxis that had become immobilized during a significant power outage in December. At the gathering, city officials voiced concerns that the immobilized autonomous vehicles hindered or redirected first responders from their primary duties.

“What has begun to occur is that our public safety officers and responders are having to physically move [Waymos],” Mary Ellen Carroll, the executive director of San Francisco’s Department of Emergency Management, stated at the hearing. “In a sense, they’re transforming into a default roadside assistance for these vehicles, which we don’t believe is sustainable.” 

Waymo informed TechCrunch that its roadside assistance personnel managed to clear numerous immobilized robotaxis during the blackout, with some still requiring to be transported by first responders.

“Waymo Roadside Assistance is a specialized team that provides additional on-the-ground support to our fleet,” the company stated in an email to TechCrunch. “Waymo’s standards for roadside response and service quality emphasize minimizing potential impacts on the community.” 

The company did not respond to TechCrunch’s inquiries regarding the number of roadside assistance employees it has or which third-party entities might employ them. Waymo also did not clarify how it intends to scale this team as it strives to launch in approximately 20 additional cities this year, broadening its reach beyond existing markets of Atlanta, Austin, Los Angeles, Dallas, Houston, Miami, Orlando, Phoenix, San Antonio, and the San Francisco Bay Area.  

Waymo’s helpers 

Waymo’s robotaxis offer over 400,000 paid rides weekly, underscoring the company’s extensive years of research in self-driving technology. Nevertheless, these robotaxis occasionally depend on human intervention, and they do so in various ways.  

The robotaxis often require guidance in complicated scenarios, especially since — as Waymo asserts — the company aims to be exceedingly careful as it expands its service.  

This guidance is provided to Waymo’s robotaxis by the “remote assistance” personnel. At any moment, about 70 of these workers are overseeing Waymo’s fleet of around 3,000 vehicles, as stated by the company. Half of these personnel are located in the U.S., while the other half are situated in the Philippines.

These specifics, which were disclosed in a letter to Congress in February, prompted critical feedback for Waymo regarding concerns over safety and security. The company has defended its use of remote assistance workers, claiming they are well-qualified and asserting that there is no significant delay caused by their geographical distance, whether they are in Arizona, Michigan, or the Philippines.  

“Our vehicle-to-RA connection is as quick as the blink of an eye. Average one-way latency is around 150 milliseconds for U.S.-based operations centers and 250 milliseconds for RAs located abroad,” the company recently conveyed.  

Remote assistance workers carry out a variety of tasks. If a Waymo vehicle encounters a challenging real-world scenario, it might issue a request to these employees to help determine the best route forward. Waymo clarifies that these workers “provide advice and support to the [robotaxis] but do not directly control, steer, or drive the vehicle.” They additionally address lesser priority inquiries from Waymo robotaxis, such as confirming if the car’s interior is clean.

However, this system isn’t without flaws.  

The National Transportation Safety Board recently disclosed that in January, a Waymo in Austin sought confirmation from a remote assistance worker about whether a nearby school bus was in the process of loading or unloading children. Although the stop sign and flashing lights were activated, the remote assistance worker mistakenly instructed the robotaxi to proceed. Consequently, the Waymo drove past the school bus while it was loading children, even as the bus’s “stop arms” remained extended, the NTSB reported.

Waymo informed TechCrunch that it “regularly audits RA responses, including their accuracy. If an incident occurs, it will be promptly flagged for follow-up actions, ranging from additional coaching to complete decertification.”

In the event of a collision or an emergency situation, Waymo relies on its “event response team.” The company asserts that this team is “exclusively situated in the U.S.” — although they are still remote — and that they are “qualified for more intricate tasks such as coordinating with emergency responders and managing post-collision procedures.”  

By that definition, the remote assistance worker who aided CHP in relocating the Waymo robotaxi from the Redwood City incident was likely part of that event response team, although Waymo did not confirm this.  

There are also growing pains evident here. Audio records from CHP dispatch, along with the incident report procured by TechCrunch, reveal that officers believed for roughly 10 minutes that Waymo intended for the passenger to drive the robotaxi away from the blaze.  

It was only when the remote worker placed a second call to 911 that CHP understood an officer was required to drive it away from the situation. (Waymo opted not to respond to specific inquiries about this miscommunication. The company maintained that it never requests riders to take control of its vehicles.) 

Next is the roadside assistance team. These individuals manage “on-site, direct interaction” tasks and are frequently responsible for relocating a vehicle. Waymo refrained from addressing inquiries about how many times these workers have moved a robotaxi, how many are available on call at any given time, or how many are situated in each city.  

Some are believed to be employed by Transdev, a third-party contractor that Waymo has utilized previously, and a few even have backgrounds as safety drivers or monitors for Waymo, based on profile data on LinkedIn.  

The company also informed TechCrunch that it “requires local tow partners to sustain rapid response capabilities for urgent towing requests and strategically place support across our service locations.” 

“Should a Waymo vehicle require assistance, we dispatch Waymo Roadside Assistance and/or local tow partners to provide on-scene support,” the company stated in a statement. “While we don’t anticipate first responders to move our vehicles routinely, we acknowledge that timing is critical in emergencies. Therefore, we established a simple process that enables first responders to gain control of the vehicle within seconds.” 

Relying on first responders  

Although Waymo claims it doesn’t foresee first responders needing to engage with its vehicles, this continues to happen — and it remains uncertain if it will be entirely avoidable.  

In at least six instances over recent months, first responders have had to manually maneuver Waymo vehicles, including at two active crime scenes.

Earlier this month, a police officer in Austin had to reposition a Waymo to allow an ambulance access in response to a mass shooting incident. In February, a first responder in Atlanta was required to disengage a Waymo after it drove into an active crime scene, before a member of the company’s roadside assistance team “retrieved it,” according to the company. This week, a Nashville police officer had to manually drive a Waymo robotaxi away after it became stuck at an intersection. 

During the March 2 hearing in San Francisco, city officials persistently questioned Waymo about what actions it would take to reduce its reliance on first responders. Waymo did not mention its personnel dedicated to moving vehicles during the three-hour session.  

District supervisor Bilal Mahmood, who supervised the hearing, expressed to TechCrunch that he felt Waymo didn’t offer many satisfactory responses. 

“I kept asking: How are you going to take more responsibility to ensure that our first responders aren’t doing that?” he stated. “And we did not receive that answer during the hearing which we were seeking, that is: What measures will they implement to ensure they take more ownership of that roadside assistance aspect?” 

A manager from Waymo’s incident response team, Sam Cooper, indicated at the hearing that the company has instructed “over 30,000 first responders globally on how to interact” with its robotaxis. He also lauded Waymo’s partnership with first responders in the creation of the system that permits them to gain control.  

“We merely want to empower them, in that scenario, to effectively relocate that vehicle from the site and secure the area so they can perform their duties,” he articulated. 

Cooper stated that Waymo has made “enhancements to our surge staffing capabilities” to be better equipped for larger emergency situations. However, he did not elaborate on those improvements, and Mahmood informed TechCrunch that his office has not received a promised follow-up. 

Cooper also noted that Waymo would contemplate leveraging collaborations similar to its engagement with DoorDash, which involves gig workers shutting robotaxi doors that were left ajar, to relocate vehicles.  

How that would differ from the existing roadside assistance staff Waymo employs remains unclear. However, city officials reiterated the same sentiment. “Our first responders should not function as AAA,” district supervisor Alan Wong remarked. 

This article was originally published March 25, 2026 at 9:30 a.m. PT.

Had Google’s AI researchers possessed a sense of humor, they might have dubbed TurboQuant, the newly unveiled, highly efficient AI memory compression algorithm revealed on Tuesday, “Pied Piper” — at least that’s what the online community speculates.

This jest alludes to the fictional startup Pied Piper, which was central to HBO’s “Silicon Valley” series that aired from 2014 to 2019.

The show depicted the startup’s founders as they maneuvered through the tech landscape, grappling with obstacles such as rivalry from larger corporations, securing funding, addressing technological and product challenges, and even (much to our amusement) impressing judges at a fictional iteration of TechCrunch Disrupt.

Pied Piper’s landmark technology in the series was a compression algorithm that significantly minimized file sizes with near-lossless compression. Google Research’s TurboQuant similarly focuses on extreme compression without sacrificing quality, but directed towards a critical limitation in AI systems. Thus, the parallels.

Google Research referred to the technology as an innovative approach to reduce AI’s operational memory without compromising performance. The compression technique, which employs a variant of vector quantization to alleviate cache bottlenecks in AI processing, essentially enables AI to retain more information while occupying less space and ensuring accuracy, according to the researchers.

They intend to showcase their discoveries at the ICLR 2026 conference next month, alongside two methods facilitating this compression: the quantization technique PolarQuant and a training and optimization strategy referred to as QJL.

While grasping the mathematics involved may be within the reach of researchers and computer scientists, the outcomes are generating excitement across the broader tech industry.

If realized in practice, TurboQuant could lower the operational costs of AI by lessening its runtime “working memory” — known as the KV cache — by “at least 6x.”

Some, including Cloudflare CEO Matthew Prince, are even dubbing this Google’s DeepSeek moment — referencing the efficiency improvements inspired by the Chinese AI model, which was trained at a much lower cost compared to its competitors on inferior chips while remaining efficacious in its results.

However, it’s important to highlight that TurboQuant has not yet been widely adopted; it remains a laboratory breakthrough at this point.

This makes comparisons with something like DeepSeek or even the fictional Pied Piper more complex. In the series, Pied Piper’s technology was poised to dramatically alter computing paradigms. In contrast, TurboQuant may result in efficiency improvements and systems needing less memory during inference. However, it does not necessarily address the broader RAM shortages associated with AI, as it exclusively focuses on inference memory, not training — which continues to demand substantial amounts of RAM.

During a press event at the White House on Wednesday, First Lady Melania Trump made an appearance alongside a humanoid robot created by the robotics company Figure AI. The pair gracefully walked down a red carpet before the robot delivered a short address, stating: “I am thankful to be included in this groundbreaking initiative to empower children with technology and education.”

Shortly after these comments, the robot casually exited the room and vanished.

This peculiar event was part of the first lady’s newly introduced initiative, the Fostering the Future Together global summit, which gathered global leaders to explore ways to enhance children’s education through technology, including AI.

The gathering certainly invoked dystopian images of the future — where the gentle (human) educator has been supplanted by a Terminator-like machine capable of walking and conversing in Latin. Indeed, during her address, the first lady urged participants to envision a future where a humanoid robot serves as the ultimate teacher for the children of the world. (The summit coincided with the Trump administration unveiling a separate tech council filled with traveling Silicon Valley leaders.)

“Envision a humanoid teacher named Plato,” the first lady declared. “Access to classical studies is now immediate — literature, science, art, philosophy, mathematics, and history — Humanity’s complete body of knowledge is accessible from your own home. Plato will offer a tailored experience, focusing on each student’s requirements. Plato is perpetually patient and always on hand. Understandably, our children will cultivate more profound critical thinking and autonomous reasoning skills.” 

“Grateful to have been invited to the White House by First Lady Melania Trump,” the Figure AI X account tweeted on Wednesday.

The first lady’s statements are clearly forward-thinking and do not accurately represent the current state or near-future potential of robotics and educational technology. Nonetheless, the notion that AI and technology can facilitate automated learning (and, in numerous ways, substitute human instructors) has been gaining traction within the tech sector. Such concepts have been consistently endorsed by the White House.

In the last year, educational innovations like the Alpha School, a series of private institutions leveraging AI to rapidly teach children, have gained momentum and attracted media attention.

The Trump administration has welcomed such experiments while simultaneously criticizing the traditional public education system.

Secretary of Education Linda E. McMahon, who is currently dissolving the very agency she leads, has still managed to visit an Alpha School campus, where she commended the “opportunity” offered by this educational network.

“Alpha School is reinventing K–12 education by equipping students with essential AI skills and preparing them for an ever-changing technology-focused job market,” the administration recently stated regarding McMahon’s visit.

Melania Trump’s event on Wednesday similarly underscored the role the administration believes the tech industry should have in shaping the future of education in America — with Trump acknowledging the “involvement of leading American technology firms, whose participation reflects the increasing importance of the private sector in fostering safe and effective educational innovation.”