分类： technology

A recent technical malfunction has left multiple Waymo driverless vehicles stranded in an Atlanta suburban neighborhood, shining a new spotlight on the ongoing challenges of scaling autonomous vehicle technology for real-world conditions.

The Alphabet-owned self-driving car firm confirmed this week that the AI-powered vehicles experienced an unexpected routing error that forced the fleet into an endless loop. Each affected car repeatedly redirected itself back to the same quiet cul-de-sac, leaving the autonomous vehicles unable to navigate out of the area on their own and requiring manual intervention from Waymo’s technical teams to resolve the issue.

Local residents reported seeing the unoccupied driverless cars circling the small residential street multiple times before the problem was fixed, with images of the stuck vehicles circulating quickly on local social media channels. Waymo has not yet disclosed how many vehicles were affected by the glitch, nor has it released details on whether the incident caused any traffic disruptions or property damage in the area.

The incident comes as Waymo continues expanding its autonomous ride-hailing services across multiple U.S. cities, including recent launches in suburban and urban markets outside of its original testing hubs. Routing and navigation remain among the most critical technical hurdles for fully autonomous vehicles, which rely on a combination of AI algorithms, real-time sensor data, and pre-mapped infrastructure to make split-second driving decisions. Industry analysts note that even rare glitches like this highlight the iterative nature of self-driving technology development, as companies work to address edge cases that do not appear during controlled testing. Waymo has stated that it is already investigating the root cause of the routing error to prevent similar malfunctions from occurring in future deployments.

Two of the technology industry’s most recognizable and influential figures are currently facing off in a landmark legal battle whose outcome could reshape the future of OpenAI—one of the most valuable startups on the planet, creator of the globally used ChatGPT. The case has already put both Elon Musk, one of OpenAI’s original co-founders, and Sam Altman, OpenAI’s current chief executive, on the line professionally and reputationally, with closing arguments now complete and the jury having retired to deliberate. Musk’s core allegation is that Altman betrayed an original agreement to keep OpenAI as a non-profit entity and effectively “stole” the organization, costing him a massive potential fortune, a claim that Altman has repeatedly and categorically denied. Over three weeks of proceedings in a California federal courtroom, journalists from across the globe have packed the benches to follow every development, with evidence ranging from incendiary private text messages to claims of free Tesla vehicles offered as quid pro quo for favorable treatment. Presiding over the proceedings is a firm, no-nonsense judge who will ultimately have final say over the ruling after reviewing the jury’s recommendation. For those unable to follow every twist of the high-profile case, here are five of the most notable insights that have emerged from the trial.

First, the vast majority of high-profile witnesses have contradicted Musk’s core claim. Musk’s entire lawsuit hinges on the assertion that Altman deceived him by abandoning the original commitment to keep OpenAI a non-profit. What has become clear over the trial, however, is that this is far from a simple he-said-she-said dispute between two billionaires. A long roster of witnesses—many of whom are among the biggest names in global tech—have testified that they never saw or heard any evidence confirming the binding non-profit commitment Musk alleges. These witnesses include OpenAI co-founder Ilya Sutskever, former OpenAI board member Tasha McCauley, and Microsoft CEO Satya Nadella, whose company invested billions of dollars in OpenAI after completing extensive due diligence. Nadella and Microsoft are named as co-defendants in the case, accused of aiding Altman’s alleged scheme, making the uniform pushback against Musk’s claims from the stand all the more notable.

Second, questions about Sam Altman’s personal trustworthiness have remained a central point of scrutiny throughout the proceedings, even with his army of high-profile supporters. In the lead-up to the trial, a bombshell investigative profile in The New Yorker by Pulitzer Prize-winning reporter Ronan Farrow painted Altman as a compulsive liar, focusing heavily on his turbulent career including his dramatic temporary ousting from OpenAI in 2023. Musk’s legal team leaned heavily into this narrative during cross-examination, opening their questioning of Altman with a blunt query: “Are you completely trustworthy?” After Altman initially responded “I believe so,” his cross-examiner pressed him on the ambiguous answer, forcing Altman to revise his response to a direct yes. Even with that correction, Altman’s character remained under intense scrutiny for the entirety of the trial. Former OpenAI board members and executives shared first-hand accounts of alleged lack of candor from Altman, and the court also learned of extensive hidden private investments Altman holds in startups that have struck commercial deals with OpenAI. One particularly controversial example is a power purchase agreement with Helion Energy, a nuclear fusion startup that has yet to deliver any operational power to the grid; Altman until recently served as Helion’s board chairman and holds a stake in the company valued at more than $1.5 billion.

Third, the trial has been defined by memorable, out-of-the-spotlight personalities that have shaped its day-to-day drama. Presiding Judge Gonzalez Rogers has emerged as a clear commanding presence in the courtroom, enforcing a strict daily schedule with just two 20-minute breaks and no lunch break to keep all participants focused. She has repeatedly and publicly called out anyone who violates court rules, from spectators who attempt to photograph the high-profile defendants to lawyers who push questioning into topics she has already ruled off-limits. Despite her strict approach, she has shown moments of dry wit: when the court experienced early technical audio issues, she joked to the room, “What can I tell you? We are funded by the federal government.” Since cameras and live streaming were not permitted in the courtroom, the public’s understanding of the trial’s atmosphere has come largely from the work of courtroom sketch artist Vicki Behringer, who has carefully captured each day’s proceedings in vivid watercolor.

Fourth, the trial has laid bare the deeply personal decay of what was once a close, admiring relationship between Musk and Altman, as well as unusual personal ties to OpenAI’s board. Musk was once Altman’s professional hero, but the relationship between the two men has deteriorated dramatically, a rift that was put on full display during testimony. When Musk took the stand as the first witness, he was largely confident and combative, but became visibly flustered when questioned about his relationship with Shivon Zilis, a Neuralink executive and former OpenAI board member. Musk confirmed to the court that Zilis is the mother of four of his children, and the pair live together. Zilis testified that Musk offered her his sperm when she expressed a desire to have children, an unusual interaction that she did not disclose to her OpenAI colleagues until a media report was imminent. Zilis left the OpenAI board shortly after Musk launched xAI, his own artificial intelligence startup competing directly with OpenAI, writing in a private text, “When the father of your babies starts a competitive effort and will recruit out of OpenAI, there is nothing to be done.”

Fifth and finally, the trial has pulled back the curtain on how power and influence operate behind closed doors in Silicon Valley, exposing unseemly backroom deals and personal power struggles that rarely see the light of day. The proceedings have revealed alleged practices that have long been rumored but never confirmed in a public court: Musk is accused of offering free Teslas to co-founders as a way to lowball their equity stakes, while Altman stands accused of making off-the-books side payments to secure key strategic loyalty. Musk’s legal team has attempted to frame Altman as a leader who leveraged his early connection to Musk to build his own personal power and influence, while Altman has countered that Musk once suggested OpenAI should eventually be controlled by Musk’s children. Private text messages introduced as evidence have laid bare the raw chaos of internal power struggles, including Altman’s frantic response to his 2023 ousting, where he asked a former colleague “still don’t want me?” The colleague responded in a text dismissing the interim CEO Altman was replaced with, Twitch co-founder Emmett Shear, as a “rando Twitch guy.” These casual, unguarded messages and the everyday sight of these billionaire icons grabbing lattes outside the courtroom have made these larger-than-life figures feel surprisingly ordinary—yet it is critical to remember that the two men still control artificial intelligence technology that shapes the daily lives of billions of people around the globe, and the dispute at hand is worth billions of dollars. Now, the decision moves first to the jury, before ultimately landing back with Judge Gonzalez Rogers to determine the final outcome.

The emerging market for AI-powered smart glasses is facing growing scrutiny over deep-seated privacy risks, even as some of the world’s largest technology corporations prepare to scale up sales of the wearable devices in the coming years, with projections pointing to hundreds of millions of units sold globally by the end of the decade.

Currently, Meta dominates the fast-growing segment, holding an estimated 80% of all global smart glasses sales thanks to its early entry into the mainstream market with the Meta Ray-Ban smart glasses, produced in partnership with eyewear giant EssilorLuxottica. The product blends the iconic, understated design of classic Ray-Ban frames with cutting-edge tech: a nearly undetectable built-in camera, integrated small speakers, and lightweight lenses that display contextual information for wearers. Recording can be activated with a simple tap on the frame, making it extraordinarily easy to capture photos and video without drawing attention to the act.

This discreet recording capability has already spawned widespread misuse, most notably a growing trend where male wearers of Meta’s smart glasses secretly record unsuspecting women in public spaces — from beaches to retail stores — for online content. Many women only discover they have been filmed after the videos, which often include harassment and abusive commentary, go viral on social media. Legal options for affected people remain limited, as photography in public spaces is widely protected under law in most jurisdictions. One woman who spoke to the BBC reported that when she requested the removal of a non-consensual recording of her, the creator demanded payment to take it down.

Beyond non-consensual public recording, the devices have faced legal pushback over the processing of user content. Two class-action lawsuits were filed against Meta after contract workers in Kenya, hired to review smart glasses recordings to train Meta’s AI models, revealed they were forced to view deeply graphic content including sexual activity and private bathroom use. Many plaintiffs said they had no idea their recordings were being reviewed by human staff, despite Meta’s assertion that the possibility of human review is clearly disclosed in its terms of service.

Even with these mounting controversies, sales have surged to more than 7 million units sold to date, with Meta CEO Mark Zuckerberg calling the product line “some of the fastest-growing consumer electronics in history” in a recent statement. Meta spokesperson Tracy Clayton defended the company’s position, noting that the firm has dedicated teams working to mitigate misuse, but argued that “the onus is ultimately on individual people to not actively exploit” any technology.

The market’s growth is attracting new entrants, with a wave of major tech companies preparing to launch their own smart glasses offerings in the next 18 months. Apple is reportedly developing its own AI and AR-enabled smart glasses, expected to launch as early as 2025. Snap has confirmed it will release a new iteration of its Specs smart glasses later this year, and Google is preparing for a second attempt at consumer smart glasses more than a decade after the high-profile failure of Google Glass, which was pulled from the market over identical privacy concerns just two years after launch. All upcoming models are expected to follow Meta’s blueprint, combining AI features with augmented reality that relies on built-in cameras to function.

Not all user experiences are negative, however. Mark Smith, a tech-savvy early adopter and enterprise software advisor at ISH, wears his Meta Ray-Bans daily. He praised the device’s practical perks: listening to music and podcasts while doing household chores without blocking out ambient noise the way over-ear headphones do, seamless hands-free phone calls, and quick, convenient photo captures while traveling that eliminate the need to pull out a smartphone. Even so, Smith acknowledged that obvious privacy risks remain, noting that the small indicator light that activates during recording is nearly invisible in bright daylight, leaving most bystanders unaware they are being filmed.

Industry projections indicate that if current growth trends hold, as many as 100 million consumers could own a pair of AI smart glasses within the next few years. That scale of adoption would create unprecedented challenges for regulators and public institutions that restrict recording in sensitive spaces including courthouses, hospitals, movie theaters, museums, and bathrooms. David Kessler, head of U.S. privacy practice at global law firm Norton Rose Fulbright, said corporate clients are already scrambling to adapt to the new reality. “There are some pretty dark places we could go here,” Kessler said. “I’m not anti-technology in any sense, but as a societal matter…will I need to think [of being recorded] anytime I go out in public?”

Risks could grow even more acute: Meta is reportedly planning to add facial recognition technology to a future update of its smart glasses, which would give wearers the ability to not only secretly record passersby, but also instantly identify them by name. Meta markets its current glasses under the tagline “Designed for privacy, controlled by you,” and advises users to avoid recording people who object and to turn off devices entirely in sensitive spaces. But these voluntary guidelines are widely ignored, with a rising trend of prank content that targets unsuspecting members of the public for online clout.

Public backlash is already starting to build. When a man posted a video complaining that a woman on the New York City Subway broke his Meta glasses after he secretly recorded her, social media users widely celebrated the woman’s actions, calling her a hero. Meta CTO Andrew Bosworth argued earlier this month that strong sales indicate broad public acceptance of the devices, but former Meta AI researcher David Harris, now a professor at UC Berkeley and AI policy advisor to the U.S. and EU, said the current generation of smart glasses faces the same fate that doomed Google Glass over a decade ago. “Technology like this is fundamentally an invasion of privacy and it’s really going to face more and more backlash,” he said.

In explosive testimony delivered Tuesday to a federal jury in Oakland, California, OpenAI chief executive Sam Altman laid out new details of a years-long power struggle with Tesla and X owner Elon Musk, revealing the billionaire’s far-reaching bid to seize total control of the ChatGPT-developing AI firm shortly after its founding — including a plan to pass that control to his children upon his death.

The courtroom appearance comes as part of an ongoing lawsuit filed by Musk against Altman and OpenAI, in which the billionaire accuses Altman of abandoning the organization’s original non-profit charter to “loot” the non-profit entity for personal gain. But Altman’s testimony flipped the narrative, arguing that Musk himself was the driving force behind restructuring OpenAI into a for-profit enterprise, all while pushing aggressively to claim full control of the company.

Recalling tense early discussions among OpenAI’s founding team, Altman described a striking exchange that he called “hair-raising.” When his co-founders asked Musk what would happen to control of the company if Musk passed away, Altman said Musk responded that leadership should transfer to his children.

According to Altman, Musk floated multiple concrete proposals to accumulate more power at OpenAI in the years after its 2015 founding. Beyond seeking additional board seats and the top executive role as CEO, Musk even proposed folding OpenAI into Tesla as a wholly owned subsidiary. The core motivation behind this push for a structural overhaul, Altman told the jury, was to access larger amounts of capital at a faster pace than the non-profit structure allowed.

Altman recalled Musk arguing that his high-profile status as a tech billionaire made him the only suitable leader for the organization. Musk allegedly claimed that a single public tweet from him would instantly generate massive value for OpenAI and attract the substantial financial backing the company needed to advance its research.

But Altman and his fellow co-founders Greg Brockman and Ilya Sutskever rejected Musk’s overtures, saying that concentrating full control of OpenAI in a single individual ran directly counter to the organization’s core mission. OpenAI was founded with the explicit goal of developing artificial general intelligence (AGI) — a broadly defined AI system that can outperform humans on nearly all economically valuable tasks — safely and for the public good, Altman noted, and the founding team never believed one person should hold unilateral power over AGI development.

“I was extremely uncomfortable with it,” Altman told the jury. “One of the reasons we started OpenAI was because we didn’t think any one person should be in control of AGI.”

The standoff ultimately led Musk to sever ties with OpenAI in early 2018. He resigned from the company’s board and halted his regular $5 million quarterly donations to the research effort. Altman told the court a blunt email from Musk announcing his exit remains “burned into my memory”: Musk insisted OpenAI had “a zero percent chance, not a one percent chance, of success” without his leadership.

Years later, when OpenAI established its for-profit subsidiary in 2019, Altman extended an offer for Musk to take an ownership stake in the new entity. Altman said Musk declined the offer outright, explaining that he would never invest in a startup that he did not control full stop. The ongoing legal battle between the two high-profile tech leaders centers on the future of one of the world’s most valuable AI companies, and the question of whether OpenAI violated its founding non-profit commitments in its push to commercialize generative AI technology.

Last week, a high-profile cyberattack against Instructure, the developer of the widely used learning management system Canvas, upended academic operations at thousands of post-secondary institutions across North America, Europe and Oceania. In the wake of the breach that disrupted mid-semester exams and locked students out of critical learning resources, the company has confirmed it reached a confidential agreement with the extortion group behind the attack to prevent the public release of 3.5 terabytes of stolen institutional and student data.

The breach, first detected on April 29, was immediately claimed by Shiny Hunters, a prolific English-speaking cybercriminal group with a track record of high-profile breaches against major global brands including Jaguar Land Rover and Gucci. The attack took Canvas offline for thousands of users, bringing exam schedules to a halt at an estimated 9,000 institutions across the United States, Canada, Australia and the United Kingdom. For many students, the disruption came at the worst possible moment: Aubrey Palmer, a meteorology student at Mississippi State University, told reporters the ransom note popped up on their screen immediately after they finished writing a 2,900-word final exam essay. Palmer and dozens of their classmates were left confused for hours, unsure if their work had been saved, before the university postponed affected exams to let students recover lost progress.

In a public statement posted to its website, Instructure confirmed the deal with Shiny Hunters, saying the hackers have formally agreed to delete all stolen data and pledged not to target individual students or affected institutions with separate extortion attempts. Under the terms of the agreement, all stolen data has been returned to Instructure, and the company has received digital confirmation that the information has been destroyed. The deal covers all customers impacted by the breach, and no individual users will need to interact directly with the criminal group, the company added.

While neither Instructure nor Shiny Hunters has explicitly confirmed that a ransom payment was exchanged, industry observers note that extortion groups like Shiny Hunters uniformly operate on a model of demanding bitcoin payments via encrypted chat platforms after successful data breaches. It remains rare for victim organizations to publicly acknowledge paying ransoms, but Instructure has opted for unusual transparency throughout the incident, updating the public regularly on its website. Analysts say this openness is likely a response to the high visibility of the attack, which directly impacted thousands of students sitting for high-stakes exams.

Instructure defended its decision to reach an agreement with the hackers, noting that protecting the personal data of students and education staff was its top priority. “While there is never complete certainty when dealing with cyber criminals, we believe it was important to take every step within our control to give customers additional peace of mind, to the extent possible,” the company said.

The decision to pay runs directly counter to longstanding guidance from law enforcement agencies around the world, which warn that paying ransoms emboldens criminal groups to carry out future attacks and provides no guarantee that stolen data will actually be destroyed. History is rife with examples of cybercriminals accepting ransom payments but retaining copies of stolen data to sell on underground black markets: when the UK’s National Crime Agency dismantled the notorious LockBit ransomware syndicate, investigators found that thousands of stolen records were still held by the group even after victims had paid to have the data destroyed.

Further context around the attack has revealed Shiny Hunters had targeted Instructure multiple times before the April 29 incident. In an interview with the BBC via Telegram, the group claimed it had hacked the company twice previously: Instructure publicly disclosed one breach in September 2025, while Shiny Hunters says it carried out an additional unreported breach in early April 2026. When asked about the widespread stress and academic disruption the attack caused for students, the group declined to comment, and also refused to disclose the size of the payment it received from Instructure.