The social media giant Meta just added a major safety feature to its digital assistant. When younger users have conversations that involve thoughts of suicide or hurting themselves, the company will step in. Instead of just showing crisis hotlines on the screen, the system will now send a direct notification right to a linked parent […]
If you want an artificial intelligence tool to book a flight or buy an audiobook, it usually needs to log into your personal accounts. Sharing your private login details directly with a chatbot feels risky, but a new software update changes that process entirely. The popular password manager 1Password just built a clever way for […]
The European Commission just told Google to stop giving its own artificial intelligence assistant special treatment on smartphones. Regulators declared that the company must give competing software the same system permissions currently used by Gemini. This order falls under the Digital Markets Act, forcing the tech giant to open up deep phone controls that were […]
OpenAI recently teamed up with Work Louder to release a physical tool for developers. Reports show OpenAI is selling $230 Codex Micro hardware product units on its website now. The keyboard brings your digital agent workspace straight to your desk. It helps users manage active chats and keep track of tasks through live lighting feedback. […]
If you have an iPhone, the AI music maker Suno just updated its software to make it much easier to share custom tracks with your friends. With a recent software update, Suno now lets iPhone users generate songs directly in iMessage without ever having to leave the chat window. You can easily turn your everyday […]
“The Lord of the Rings: The Hunt for Gollum,” will de-age actors using AI, the actor and director Andy Serkis confirmed to Variety.
“There’s a little bit of de-aging for some of the characters and machine learning is part of the process,” Serkis, who is both directing the new film and reprising his famous role as Gollum, told Variety.
AI is an incredibly contentious topic in Hollywood. Some studios, like A24, have leaned into generative AI. Others in the industry, from actors to directors and writers, have lambasted the tech, with many fearing that AI will be used to replace industry jobs or simply result in lower-quality work. As the tension between the pro-AI and anti-AI — or perhaps AI-resigned — factions deepen, the industry and its leaders, Serkis included, are publicly grappling with the technology and what it means for a key creative industry.
In addition to Serkis, returning actors from past films include Elijah Wood as Frodo, Ian McKellen as Gandalf, and Lee Pace as Thranduil — all of whom are markedly older than when the original “Lord of the Rings” trilogy and the prequel franchise “The Hobbit” were filmed. Beyond de-aging, though, Serkis said that “at present,” there are no other plans to use AI in the making of the new movie.
“We’re not creating AI shots in our movie, every shot is created in a traditional way,” said Serkis. “One of the things actually that I really wanted to do with this film was to bring back all of the great filmmaking skills, from miniatures to prosthetics and marry them up, because that’s my taste.”
Serkis didn’t nix the idea of using AI altogether. He noted that Peter Jackson, who famously directed the previous live-action movies in the celebrated franchise, worked with the software engineer Stephen Regelous to develop the AI-powered special effects software known as MASSIVE back in the early 2000s.
“When you think about it, in the original ‘Lord of the Rings’ films, Peter created MASSIVE, which was a program which allowed thousands of orcs to all have their own individual mindset,” said Serkis. “So that is a brilliant example of an incredible use of AI.”
He added that AI can be a “valuable” creative tool “as long as it’s not exploitative and as long as it’s not harming anybody or defaming anybody or telling mistruths.”
“When it becomes exploitative and people are not remunerated for the work that they’ve done, or it’s used in nefarious, or mean-spirited or pornographic ways, then of course, that’s terrible and we’ve only brought it upon ourselves,” he added. “I’ve said this before and I mean it — we are the parents of AI, and we have to be good parents and teach AI well. And if we teach AI well, then it can help us across many industries.”
As The Tab noted, many Lord of the Rings fans responded positively, drawing a line between machine learning-infused virtual effects software and generative AI tools that churn out text, audio, and imagery.
“AI is perfect for de-aging,” wrote one commenter. “It should get a complete and utter free pass for that. Having an artist sit and de-age someone must be a horrific chore. Let them work on something fun instead.”
“Every single VFX tool these days uses AI for some part of the processing,” added another. “This isn’t generative AI.”
Meta executives claimed that a controversial facial recognition feature didn’t “exist.” Fast forward a few weeks later, and they’re describing the feature in public — and in detail.
In June, a Wired investigation revealed that Meta had quietly infused unreleased facial recognition tech into its “AI Glasses,” the tech giant’s chatbot and camera-equipped smart glasses. While the feature, dubbed NameTag, was inaccessible to consumers, it was designed to register faces encountered by Meta glasses wearers “into unique biometric signatures, commonly known as faceprints, and check each one against faceprints stored on the user’s phone,” according to the magazine.
Andy Stone, the tech giant’s spokesperson and VP of communications, declared on X that Wired’s reporting was “intellectually dishonest” and “advocacy-driven click bait.” What’s more, according to Stone, NameTag was fully nonexistent.
“Here’s a thing: Wired reports Meta didn’t answer several questions about how this will work,” Stone added in a follow-up post. “How could we? The feature doesn’t exist!”
In response to Stone, another Meta executive, chief technology officer Andrew “Boz” Bosworth, jumped in.
“Incredibly misleading from Wired,” he wrote. “Absolutely dishonest.”
Yet despite company leaders reassuring us that the feature doesn’t exist, Bosworth discussed it in detail during an interview with The Atlantic CEO and podcaster Nick Thompson. Bosworth’s comments in the interview, which was published last week, seemingly contradicted his fellow executives, revealing that the feature did indeed exist.
NameTag, Bosworth told Thompson, would be “encrypted locally to your device,” and would catalog people who smart glasses wearers “met in person, with your glasses on, who introduced themselves, or you said, ‘okay, this is David, remember this person.'”
This information is “only available to you, when you are wearing your glasses,” Bosworth added. He emphasized that the data captured by NameTag wouldn’t be stored in a centralized database.
“This is a person you’ve met before. Here’s their name. They’re right in front of you. That’s… what we call a NameTags feature,” said Bosworth. “I think people get confused because they hear face recognition and they think, ‘oh, there’s a central face database that everyone’s being scanned constantly into.’ It’s not that.”
In addition to supporting blind or low-vision smart glasses wearers, said Bosworth, NameTag could solve what Meta calls “the cocktail party problem,” saving wearers from any awkwardness that may arise when encountering people at parties who they’re sure they’ve met before, but can’t quite place.
“It’s a very universally human problem,” Bosworth said. “Nearly every human I talk to, no matter what their status of their vision or their memory, understands that this is a problem that would help them feel more comfortable in social situations if they had it.”
To be clear, Wired never suggested that the facial recognition feature it uncovered was being used to build a “central face database” by way of Meta’s smart glasses. Indeed, in line with what Bosworth said in the podcast appearance, Wired reported that “code discreetly added to Meta’s AI app over multiple updates this year shows that the feature, internally called ‘NameTag,’ identifies people captured by the glasses’ camera and, when activated, alerts the wearer when it recognizes someone.” The story also repeatedly noted that the feature was unreleased and inaccessible to glasses wearers.
When journalists and others called out the contradiction between Meta’s past and recent statements, Stone — the comms executive who said that Wired couldn’t answer questions about the feature because it “doesn’t exist” — jumped back in the comments section.
“We have said for months we’re exploring such features, even as nothing has shipped to consumers,” the spokesperson said in response to a post by The Intercept journalist Sam Biddle. “No secret there despite the conspiratorial thinking.”
There’s a big difference between a feature not existing and a feature that exists — if still unfinished — and has yet to be released. Meta appears to be arguing that if a feature remains inaccessible to consumers, it’s technically nonexistent, even if the company has gone as far as to issue multiple updates inserting code for the feature into products that are already in consumers’ hands. (Meta quietly deleted the NameTag code after its existence was made public.)
During the podcast interview, Bosworth took care to emphasize that Meta is considering the privacy implications of a facial recognition-powered feature like NameTag. He noted that it’s likely illegal in states like Illinois and Texas, and that it “has to be done in a way that people feel comfortable.”
Meanwhile, Meta continues to fight an ongoing consumer harm lawsuit, prompted by an investigation by a pair of Swedish newspapers. The publications found that Meta collected recordings by smart glasses wearers and passed them along to human data labelers, despite promising consumers that their recordings would be stored only on their devices.
“Designed for privacy,” reads a webpage for Meta’s smart glasses, “controlled by you.”
Those environmental burdens also come with a financial one: skyrocketing demand for electricity, which is increasing the cost of utilities for renters and homeowners alike.
According to a report by Monitoring Analytics, an independent monitor for the largest transmission company in the US, PJM, data center demand is expected to drive over $23 billion in customer price increases by 2028. That eye-watering figure, first spotted by Fortune, is a direct result of the country’s old and confusing electrical infrastructure, which ultimately leaves regular people, not the tech industry, holding the bag.
Whether they’re data centers, factories, or other large facilities, Fortune points out that local regulators and transmission companies like PJM have a hard time figuring out who’s responsible for rising energy demand. While a small power line from a data center campus to a nearby substation is easily billed to the data center, figuring out who to invoice gets harder further on down the line, particularly with shared infrastructure like the substation itself, or the long-distance transmission lines connecting to it.
Though utility companies can and do charge data centers for the electricity they use, Fortune observes that the facilities undergirding the AI boom can “fine tune” their electricity use minute-by-minute, gaining an edge that the average consumer could only dream of.
Because many utility companies charge based on a system of “peak demand” — a measure of a customer’s energy usage at the exact moment the collective grid hits peak demand — data centers have gotten into the habit of scaling down right when the grid measures highest demand. That’s not because they’re actually trying to use less power, but because that narrow window is what determines how high their bill will be. In reality, their overall usage remains the same.
This is more or less the scenario that played out in a Bitcoin mining operation in Texas, where the company Riot Platforms agreed to cut its power use on hot summer days, only to ramp back up massively at night. In exchange, the company negotiated for a lower flat electricity rate overall, and even snagged some state subsidies meant to encourage responsible energy use.
Load-shifting like this can genuinely help the grid, but it also means companies with the ability to game the system get rewarded with subsides that an average household will never get. And because companies are merely changing when they suck their juice — not how — subsidized load-shifting ends up being a poor strategy for reducing electrical use overall.
And while data centers are just the most contemporary example of this practice, the shrewd manipulation of US energy markets by for-profit corporations dates back decades. That may not be comforting news for citizens scrambling to hold data center firms accountable in the here-and-now, but it should at least give them an idea of where tostart looking for inspiration.
Meta patented AI tech that constantly analyzes your voice and surroundings in order to detect your emotional state. Finally, exactly what we were all asking for: Meta being granted constant access to our offline lives.
As first caught by Patentlyze, a patent published on July 2 describes a system that records all the user’s “audible communications” and combines it with “contextual factors” like “time of day, location, user activity, or digital interaction.” Audio may be “transcribed, and an emotional-state machine learning model may interpret verbal and nonverbal cues to determine emotional indicators.”
In other words, the system listens to you and everything you do, combines it with other data, and feeds it all to a mood-predicting AI.
“The system increases the precision and reliability of emotional inference by aligning multimodal sensor inputs on synchronized timelines, which creates a novel data structure that supports richer emotional analysis,” reads the patent. “These combined features deliver a technical improvement in automated audio interpretation, enabling continuous emotional monitoring on everyday devices.”
According to the patent, the goal of this extraordinarily intrusive system would be to better tailor users’ workouts to their emotional state, as “personal trainers cannot provide the level of precision in guidance, such as correcting a pose and/or body movement,” as well as Meta’s imagined device could.
It’s not surprising to see Meta pitch this kind of surveillance technology as a fitness device. Health-tracking wearables have become wildly popular, despite the extraordinary amount of physical data about individual consumers they collect and store.
The key difference in Meta’s case, is that rather than crunch biological metrics, it’s an AI-powered fly-on-the-wall attempting to measure a person’s emotional state by listening in on the world around them, and possibly then connecting that information to all other available data.
“The AI assistant may listen to a user(s) at predefined times to hear various types of communication, such as sighs, laughter, and/or the tone(s) of a voice(s),” reads the patent. “The AI assistant may use these inputs to quantify the user’s emotional state or generate other insights about the user.”
The patent also discusses connecting audio inputs to information like when a user takes their medicine, proposing that “the AI assistant may take multiple inputs in addition to audio inputs (e.g., of a user’s voice) to provide a summary of emotional trends based on various inputs (e.g., a happier emotional state associated with a particular time of day or at a time when medication is taken, etc.)”
It’s worth noting that fellow tech giant Amazon tried and failed to market a similar system. In 2020, the company launched the Halo Band, a fitness band with a built-in microphone that was designed to listen in on users and conduct what Amazon described as a “tone of voice analysis.” Following public backlash in 2021, Amazon nuked the microphones when it launched a next-generation version of the band. It discontinued the product line altogether in 2023.
Whether we’ll ever see Meta attempt to launch a similar mood-tracking device is unclear. In a statement to 404 Media, a Meta spokesperson said that “like other companies, patents at Meta are often filed to disclose concepts that may or may not be implemented, and a granted patent does not guarantee that Meta has pursued or will pursue the technology described.”
The existence of the patent, though, shows that such a device is certainly on Meta’s mind — and is yet another Meta-made product that stands to strengthen the bridge between our online and offline worlds.
A reverse prediction test suggests that supposedly brainlike AI models may rely on visual strategies the primate brain does not use. Artificial intelligence can sometimes predict how the brain responds when people recognize objects. But that resemblance may hide an important weakness: the internal workings of today’s vision models do not necessarily match the processes [...]
A hack revealed in detail how AI music generating app Suno scraped millions of songs, likely including copyrighted ones, from across the web to feed into its AI model, 404 Media reports.
Suno, which is currently embroiled in multiple ongoing copyright lawsuits, has already admitted in response to legal action that it used “essentially all music files of reasonable quality that are accessible on the open internet” to train its music-generating AI.
According to 404, the hack shed light on how that copyrighted music was obtained and funneled into datasets, with Suno pulling music and lyrics from sites including YouTube Music, Genius, and Deezer, in addition to stock music libraries like Jamendo, Freesound, Pond5, and the International Music Score Library Project (IMSLP).
Source code obtained by the hacker also pointed to the staggering scale at which Suno has swallowed up music from across the web. For instance, one file, titled “youtube_music,” contained “2,013,545 music clips,” while another file noted that Suno’s AI datasets held “113,879 hours of youtube_music.” Altogether, the file appears to host hundreds of thousands of hours of music.
The hacker, who goes by the pseudonym ellie.191, also claimed to 404 that they’d also been able to view sensitive Suno user data, including Stripe payment information.
In a statement, Suno told 404 that “as we have stated in public filings and disclosures, Suno’s AI models have been trained on publicly available music files and related metadata accessible on third-party websites on the open Internet.”
The company also insisted that its “goal has always been to help people create original new music, not replicate someone else’s.” That’s despite tests extensively showing how Suno can easily replicate published music, including the work of well-known artists.
Yet another domino appears to be falling as part of the Rube Goldberg machine that will eventually pop the AI investment bubble.
Earlier this month, economic forecasters were sounding the alarm that overspending on AI was at a level far more severe than Black Tuesday, the day that jump-started the worst economic catastrophe in the history of the industrial economy. Now, investors seem to be coming to that same conclusion all on their own.
This week, Taiwanese semiconductor giant Taiwan Semiconductor Manufacturing (TSMC) posted its second-quarter earnings results, revealing a staggering revenue of over $40 billion — a record-breaking sum for the company.
While that should come as welcome news to investors, the results had the exact opposite effect, sending shares of TSMC stocks tumbling by four percent. That in turn led the tech-heavy Nasdaq 100 index to fall by 1.4 percent Thursday, compounding losses from Wednesday, Bloomberg reported.
The issue seems to be TSMC’s revision to its capital expenditure. As the key manufacturer for chip design company Nvidia — arguably one of the most pivotal players in the AI boom — TSMC is a major bellwether for investor confidence around the buzzy tech.
The trouble is that, in addition to posting record revenue, the Taiwanese chip firm raised its 2026 spending forecast to a range of $60-64 billion, up from $52-56 billion. That shift will test how much more spending investors are willing to stomach on AI, a technology that has yet to justify the nearly $1.6 trillion spent developing it over the past decade.
Zooming out, the market’s response seems to indicate a few things. For starters, good news is clearly not enough to maintain confidence in the tech industry’s ability to make AI a financial success. It likewise signals that belief in the AI bubble is no longer confined to a few brave contrarians, but is becoming a mainstream narrative, as investors grow wary of an industry which promises the Moon, yet only seems to deliver hot air.
Moonshot AI’s Kimi K3 offers 2.8 trillion parameters and planned open weights, raising new questions about AI costs, control, and US-China competition.
OpenAI is selling its first hardware — without any help from Jony Ive. It describes the Codex Micro as a “command center for agentic work” but it’s really a 13-switch wireless keyboard customized to help developers keep tabs on what their Codex agents are doing. It costs $230.
The keyboard has 13 mechanical switches (one keycap covers two of them by default), a rotary encoder, joystick, and RGB backlighting around the whole keypad and individual keys. It comes with 32 customizable icon keycaps.
OpenAI claims that the Codex Micro is a serious business tool: The command keys enable Codex users accept changes, reject outputs, push-to-talk, start new chats, and trigger custom actions. The rotary encoder can be used to dial up the “brainpower” allocated to tasks — or in more conventional terms, how many tokens to allocate to reasoning on a task. And the RGB lighting can provide feedback on how various tasks are progressing. And the RGB lighting under the “agent” keys can provide feedback on how various tasks are progressing.
The Codex Micro’s manufacturer, Work Louder, already has a similar device on the market, the Creator Micro, which offers similar functionality, but without the colorful keys. It costs $56 less than the Codex Micro, however.
The Codex Micro will fit in snugly with OpenAI’s other merchandise, where using Codex is as much about a fashion statement as a technological choice.
OpenAI is selling its first hardware — without any help from Jony Ive. It describes the Codex Micro as a “command center for agentic work” but it’s really a 13-switch wireless keyboard customized to help developers keep tabs on what their Codex agents are doing. It costs $230.
The keyboard has 13 mechanical switches (one keycap covers two of them by default), a rotary encoder, joystick, and RGB backlighting around the whole keypad and individual keys. It comes with 32 customizable icon keycaps.
OpenAI claims that the Codex Micro is a serious business tool: The command keys enable Codex users accept changes, reject outputs, push-to-talk, start new chats, and trigger custom actions. The rotary encoder can be used to dial up the “brainpower” allocated to tasks — or in more conventional terms, how many tokens to allocate to reasoning on a task. And the RGB lighting can provide feedback on how various tasks are progressing. And the RGB lighting under the “agent” keys can provide feedback on how various tasks are progressing.
The Codex Micro’s manufacturer, Work Louder, already has a similar device on the market, the Creator Micro, which offers similar functionality, but without the colorful keys. It costs $56 less than the Codex Micro, however.
The Codex Micro will fit in snugly with OpenAI’s other merchandise, where using Codex is as much about a fashion statement as a technological choice.
Elon Musk’s xAI is suing a man for using its flagship chatbot, Grok, to create nonconsensual sexual deepfakes of women and children.
The Grok user, a 67-year-old in South Carolina named Terry Wayne Harwood, was arrested in March on multiple counts of sexual exploitation of a minor. According to the allegations, Harwood possessed and distributed child sexual abuse material, or CSAM. Filed earlier this week in Texas, xAI’s lawsuit accuses Harwood of using Grok to do just that, per Reuters. He apparently used two different Grok accounts, according to the lawsuit, and though xAI claims that Grok refused some of Harwood’s requests, he was able to tailor prompts to sidestep Grok’s safety guardrails and create the disturbing imagery.
The lawsuit follows a massive deepfake scandal that unfolded over the winter, as X users realized that Grok — which had been updated with a new photo and video-editing feature — could be used to quickly and easily alter images to artificially undress real people and depict them in sexual positions.
As Futurism reported, some of these images depicted real women in sexually violent and degrading scenarios. Some of the nonconsensual deepfakes depicted minors. The Center for Countering Digital Hate estimated that during an 11-day period, Grok “generated an estimated 3 million sexualized images, including 23,000 of children.”
Now, as it sues one of its users for the creation of illegal content, xAI is fighting multiple lawsuits over Grok-generated CSAM. In March, teenagers in Tennessee sued the company in a proposed class action, alleging that Grok had been used to create deepfakes of them and more than a dozen other minors. Just last week, the suit was amended to include allegations that a man used Grok to generate more than 7,000 sexual deepfakes of his stepdaughter. The amended complaint alleges that the man’s prompts largely went unflagged by Grok’s guardrails.
In another high-profile case, the former conservative influencer (and mother to one of Musk’s many children) Ashley St. Clair sued xAI after she was aggressively targeted by sexualized deepfakes, including one depicting her as a minor.
Law enforcement professionals and child online safety experts say that AI has led to an overwhelming and disorienting new flood of CSAM, as Bloomberg detailed in a devastating report earlier this year. It’s also created a host of new challenges for those attempting to find and arrest child predators, as AI has made it more difficult for investigators to determine whether a child depicted in abuse material is real or AI-generated.
In its latest lawsuit, xAI accuses Harwood of engaging in a “calculated scheme to weaponize Plaintiff’s tool for criminal ends, exposing real victims to profound and lasting harm, while exposing Plaintiff to significant legal risk and reputational damage.”
xAI — which is now owned by SpaceX and has since been renamed to SpaceXAI — also alleges that it “enforces its rules against violators through account suspensions, account terminations, and by reporting suspected child sexual abuse material to the National Center for Missing and Exploited Children,” or NCMEC. It also claims that xAI “has suspended 52,222 accounts and made 73,604 reports to NCMEC in 2026, resulting in (at least) 244 arrests.”
But despite its best efforts, it’s clear that the startup still has a serious deepfake problem. If a user is engaging in deeply disturbing criminal activity using a company’s service, it’s reasonable for a company to hold them to account. But suing users doesn’t fix a chatbot’s failing guardrails, facilitating their crimes.
Login
