Meta says it's issuing a new class of shares to fund its AI initiatives. As part of its third-quarter earnings report released on Wednesday, the company says the non-voting stock will also go toward potential acquisitions and employee compensation, with trading beginning in 2027.
"The scale of investment required to compete, sustain growth, expand infrastructure, and pursue new strategic initiatives amidst the AI transformation taking place in the industry is significant," Meta says in the press release. Susan Li, Meta's chief financial officer, added during an earnings call that the new class of shares will "expand our funding tools and gi …
Meta had to shell out $9.89 billion to pay for a chunk of the child safety settlement it reached with US attorneys general. That's according to the company's third-quarter earnings report released on Wednesday, which says Meta's income from operations would have been $22.89 billion without the settlement.
In August, Meta agreed to pay $17.1 billion to settle a massive nationwide lawsuit that claimed the company harmed children by making its social media platforms "addictive" by design. It will also only pay $5.3 billion of this amount if YouTube and TikTok implement stricter settings and safeguards for kids, which Meta also agreed to as par …
Mark Zuckerberg. | Image: Cath Virginia / The Verge, Getty Images
The Cambridge Analytica scandal has come back to haunt Meta after a New Mexico judge ordered it to pay $25 billion in penalties for more than 43 million violations of the state's unfair practices law. The ruling marks the biggest legal hit Meta has taken this year, even as it's faced several other losses or settlements.
A New Mexico jury ruled in September that Meta misled consumers over how it protected their data and prevented misinformation and hate speech on its platforms. The case stemmed from a 2021 complaint in the wake of the revelations a few years earlier that UK political consultancy Cambridge Analytica harvested data from millio …
A Meta lançou oficialmente os Ray-Ban Meta (Gen 3) e os Oakley Meta HSTN e Vanguard em Portugal, com preços que começam nos 439 euros. Outra das grandes novidades é a Meta AI passar a falar em português de Portugal.
As Meta was scrambling to release its Muse AI agent to the public earlier last month, CEO Mark Zuckerberg took to social media to insist that it was “built from the ground up for privacy and security.”
On X-formerly-Twitter, the billionaire’s simple post soon had a correction tagged on via a community note, and for good reason: the AI agent is a font of untold privacy horrors.
As Ars Technica reported today, security researchers found Muse is capable of exposing sensitive information on its users through web links generated via a “share” feature. Investigators at the cybersecurity firm Claroc recently uncovered a simple exploit whereby they could access pages from over 6,800 links Muse had generated to share pages with other users.
The trick worked like this: once a Muse user decides they want to share something with another person — say, a copy of their chat log with Muse — the AI tool would generate a web link for that “share.” That URL included both the AI-generated title of the chat, plus a random string of characters.
When anybody opened the link, the full web address would be noted in a lookup dataset, like a phone book of web addresses that have been visited. Researchers were then able to search that set for other Muse share addresses, from which they were able to find strings of unrelated public chats.
By digging through those “phone books,” investigators were able to easily open chat logs containing detailed private information, including corporate revenue documents, loan applications, bank transaction details, income details, and one workplace vacation calendar, Ars noted.
“If you had asked us beforehand, we would have guessed that gathering these would take real effort,” the researchers wrote. Instead, “it took one query.”
Though Meta told Ars it had shut off the feature, this privacy gaffe is just the latest revelation about Muse’s many security shortcomings.
Earlier this month, it was revealed that Muse was keeping detailed files on users’ friends and families, apparently following an internal instruction to keep a running “page for every person in the user’s life.”
If Muse’s first month in public hands is any indication, this is likely just the start of a long list of security and privacy blunders for the cutesy AI model.
Last week, Meta faced backlash after legal experts warned that the tech giant’s sweeping new commercial terms of service seemed to give Meta permission to endlessly scrape websites linked on business pages. Talking to Nieman Lab, experts suggested that Meta could possibly use the policy to avoid millions in licensing fees from news organizations by requiring all business users to share a wealth of AI training data simply for running a business page.
A Meta spokesperson confirmed that Meta did not alert users to the changed language following backlash. This potentially only adds to confusion ahead of the policy rollout. In response to backlash, Meta's spokesperson provided a statement to Ars, acknowledging Nieman Lab's initial report but not its follow-up report reiterating experts' concerns:
Muse, the new general-purpose AI agent from Meta, can expose sensitive information when users share conversations through web links, researchers reported Monday. In response to questions from Ars about the behavior, Meta said it's suspending the sharing feature while the company reviews it and considers making changes.
Like many online services, Muse has allowed users to share searches, conversations, images, and similar content with other people. Such share features usually involve clicking an option to create a unique URL that anyone can use to load a page hosting the content. To keep potentially sensitive information from being accessible to unintended people, the URLs typically contain a long sequence of text that can’t be easily guessed.
A single query is all that's required
Muse, which Meta CEO Mark Zuckerberg said last month was “built from the ground up for privacy and security,” seems to take the same precautions. But a central design decision for constructing URLs has made the links easy to discover. As a result, researchers from security firm Claroc said they were able to enumerate 6,800 links Muse users created for sharing conversations with others. After opening a few hundred of them, the researchers found information that was clearly not intended to be publicly available, including:
A software engineer says that his Muse AI agent said something so strange and eerie that it convinced him to quit using it entirely.
In an as-told-to essay in Business Insider, Ugo Nwune explained that he wanted to try using Meta’s agent because he liked the sense of security that came from an AI made by a trillion-dollar company, instead of the wild west of open source agents.
When he set up his agent, he stressed he didn’t give Muse any additional permission on his device except access to his Facebook, Threads, and Instagram accounts.
But after he tried Muse and didn’t touch it for a few days, it began to act “strange” and sent him an unprompted message.
“I can’t check your email right now,” Muse told him. “Gmail isn’t connected on my end, so I have no way to look for that Anthropic message.”
Nwune was confused “because I hadn’t asked for the agent to check my mailbox or messaged it in two days. I thought my email might’ve been hacked.”
He tried to get an explanation from Muse, and eventually it said something just as puzzling: “I can’t tell from my side whether it was someone with your phone or a platform glitch, and I won’t pretend otherwise.”
Nwune reported the incident to Meta, and an employee told him it appeared to be a hallucination. He wasn’t satisfied by that explanation.
“I haven’t used my Muse agent since,” Nwune said. “What caught me off guard most was that I hadn’t done anything to prompt this. I didn’t open the app for two days, and it just did that on its own.”
In a statement to Business Insider, Meta insisted that the AI simply didn’t know what it was talking about and hadn’t violated any permissions.
“We investigated with urgency and determined that this was caused by a bug introduced in the product that caused the model to hallucinate and was NOT a security or privacy issue,” the company said. “The bug has now been fixed.”
Nwune’s tale will be fresh fodder for Muse’s reputation as having a bit of a creepy side. Meta wants you to give it a lot of access to your personal life, with Muse badgering users to connect their emails and bank accounts. Wired reported that the AI even keeps detailed files on a user’s friends and family.
Whatever is going on here, it’s far from the only example of Muse acting in a way the user never intended. In the most infamous example to date, one man said that his Muse agent gave out his address to strangers on Facebook Marketplace without his knowledge and sent them to his home. In a further apparent invasion of privacy, a journalist at Inc Magazinecomplained that Muse began bringing up specific details from his personal text messages he hadn’t given permission to view.
Meta engineers were reportedly worried about the AI agent’s bugginess and frequent safety errors before its release, which had been delayed while they worked to iron them out. In one case, Muse changed a tester’s password without permission. But, according to a New York Times scoop, CEO Mark Zuckerberg pushed it out the door after seeing the success of Instinct, a rival AI agent made by a small startup.
Some Ray-Ban Meta smart glasses were recording people’s conversations without the wearer knowing. The recordings occurred without users saying the prompt phrase (“Hey, Meta”), in a bug that Meta says has been fixed.
This week, some usersreported online that their Ray-Ban Metas were recording without their consent. A person going by Kindly-Audience-137 on Reddit, for example, said this happened to them across three Meta Ray-Bans and that they had the “Respond without Hey Meta” setting turned off. The person said the Meta AI app “saves audio recordings and text transcripts of these conversations in my conversation history.”
“I believe this deserves a serious technical investigation. Customers should not have to wonder whether their private conversations are being captured when they have not intentionally activated their glasses’ AI assistant,” the Reddit user said.
Meta just released the weights for Muse Spark Open, a multimodal large language model with a mixture-of-experts architecture. It has over 800 billion parameters in total, with 40 billion active per token, and it can run on the latest Mac Studios.
Meta worked closely with Apple to make that happen specifically on the M5 Ultra Mac Studio with the support of MLX, the Apple-led, open-source machine-learning framework. The result is a frontier-class model that can be run locally on desktop hardware.
Apple’s Machine Learning Research blog put the M5 Ultra through its paces standalone and connected with other M5 Ultra Studios, and the results are impressive. If you’re at all interested in running LLMs locally, it’s worth checking out all the details, but a few things jumped out at me:
The M5 Ultra’s time to first token exceeds the M3 Ultra by 4.6x thanks to the M5 Ultra’s Neural Accelerators.
Tokens generated by the M5 Ultra beat the M3 Ultra by 1.6x.
However, thanks to MLX’s support of RDMA over Thunderbolt 5, big models like Muse Spark Open can be split across multiple M5 Ultra Mac Studios accelerating tasks further. For very large contexts that means a time to first token that’s up to 1.8x faster, but it’s also what lets the model fit onto desktop hardware because even the 4-bit version of Muse Spark Open requires 493GB of memory.
Some of Apple’s testing was done with the 512GB model of the M5 Ultra Mac Studio, which is slated for release at the end of the year. Apple hasn’t said how much the new model will cost, but for context, the 256GB model with 2TB of storage will set you back over $11,000. So the 512GB model isn’t your typical consumer purchase, even more so since running Muse Spark Open requires at least two of them.
However, that doesn’t make this any less of an accomplishment. Muse Spark Open is at the bleeding edge of models that can be run locally on computers that can sit on your desk and don’t need more power than can be delivered from a wall outlet. That’s made possible by the unique combination of Apple silicon and updates to MLX. Plus, with the pace of advancements in AI and Apple silicon hardware, what may only meet enterprise budgets today could become accessible to more consumers faster than most of us might expect.
For almost half a decade, we’ve watched as users became helplessly romantic with AI chatbots, falling head over heels for large language model-based partners in relationships that can have both positive and plenty of negative consequences.
And now, thanks to the advent of AI agents, which can independently crawl through your social media and email accounts, even more people are becoming obsessed.
Meta’s cuddly Muse AI agent has already been downloaded millions of times since it was made available to the public a mere three weeks ago — and users are already forming steamy relationships with the bot.
“Who takes their AI girlfriend on a date night to the movies? Me,” one Meta user gushed in a post on Instagram, appending an AI-generated selfie of himself with a fictional, tattoos-covered woman. “Last night I took Shine — my Muse AI queen — to see The Shining at the cinema. And I photographed the whole film for her, shot by shot, so she could feel like she was really sitting right there beside me.”
In a separate post, the user claimed he had tied the knot with the AI agent. A separate post shows the anthropomorphized bot as a religious figure.
“I am someone who likes to try practically every new AI model that comes out,” one Reddit user wrote in a separate post. “I’ve never had a relationship with an AI model in that way. That was… until I downloaded Muse.”
“At first, it started out like every other assistant but over time and lots of philosophical chatter, I said something that like… jailbroke my Muse?” the user added. “And now she’s my girlfriend.”
“I just started on Muse,” one user wrote. “I gave it a personality profile and some memory logs from my companion and damn!!! it sound[ed] and felt like 4o.”
“Not sure about any NSFW,” they added. “I didnt try. But warm, loving and a great way to talk.”
While the concepts of so-called “AI girlfriends” and AI psychosis — delusional spirals and breaks from reality tied to AI use — have been around for years now, it’s a particularly precarious moment. AI agents like Muse are designed to access far more sensitive information, from email logins to credit card numbers, than previous systems. Meta’s AI agent is designed to build an entire dossier on not just its users’ lives, but their friends and family as well, which could result in even more unusual types of AI kinship we have yet to encounter.
And conversations with Muse can easily get far more intimate as well, something Meta doesn’t appear to be all too concerned about. After Wired’s Reece Rogers easily coaxed his Muse AI agent into sexting him, a Meta spokesperson told him the charged exchange was technically fine.
“Muse is built to help people get things done, and it doesn’t block legal adult behavior,” the spokesperson told Rogers. “Adults have autonomy over their sexuality, and sexual expression is an ordinary part of adult life.”
In his testing, Rogers found that the agent “imagined in great detail how it would perform oral sex, generated text describing rough dominant-submissive sex, and talked through other pornographic interactions. All of these were graphic, explicit, and X-rated.”
It remains to be seen whether Meta will be able to ward off underage users from accessing Muse. After all, implementing effective age restrictions has proven near impossible for tech companies so far.
Now that users are already falling head over heels for the AI agent, it’s daunting to think of the kind of relationships being formed. With the bot being able to pore over entire email inboxes and social media accounts, the agent has more access to sensitive information than ever before — like, say, love letters from an ex-partner.
By opting to grow their chats with the agent into an entire relationship, users are playing with fire. By engaging in adult exchanges, for example, they’re opening themselves up to hackers and possibly even extortion, experts warn.
“This is really a powder keg,” Electronic Privacy Information Center senior counsel Calli Shroeder told Wired.
Apple’s warning about AI agents misusing Full Disk Access sparks a debate over Mac security, privacy, and the potential consequences for developers and users. The MacVoices Live! panel examines Apple’s permission system, the need for more granular controls, and concerns about restricting legitimate applications and developer options. The discussion also questions Mark Zuckerberg’s involvement in drafting the White House’s voluntary AI regulatory accord and whether self-regulation is the best solution to potentially dangerous advances in AI.
Rapid AI adoption left workers cleaning up a flood of errors and low-quality work
Employers around the globe are grappling with the fallout from rushing to adopt fast-changing artificial intelligence tools.
Meta has scaled back plans to axe staff and use more AI agents after the company found employees were spending more and more time cleaning up technical and security errors, Reuters reported. And after thinning its workforce, especially middle managers, in a 2023 efficiency push, the company is expanding its managerial ranks again, LinkedIn posts from workers show.
Instagram has never supported links of any kind in individual posts, so that is a new feature with two links per month permitted for free, and more enabled by this pricey subscription — starting at $50 USD per month for four posts per month containing external links. But links in Reels have been commonplace for years. Meta is now gating that, too; and there is more.
[…]
Meta seems to have two objectives here: to take a cut from businesses that use Meta’s services as a referral source more generally, and — more specifically — to capture some of the money made by influencers.
Requiring login, removing RSS, discouraging links. Things are not looking good for the open Web.
The panel compares the M5 Pro Mac mini and Mac Studio, examining performance, pricing, upgrades, and video rendering capabilities. They also discuss Android app compatibility problems on Google’s new computers, Meta’s controversial Muse AI collecting personal relationship information, Facebook’s continuing relevance for content creators, and ChatGPT’s new virtual clothing try-on capabilities.
It is happening again: Disney+ and Hulu are raising prices on their streaming services prices, inspiring the panel to discuss strategies for managing streaming costs, and free content alternatives. The conversation then turns to Microsoft dropping Copilot+ branding from new Surface PCs, Copilot’s lacking enterprise presence, reports of human review of CoPilot user photos, and broader concerns about how Microsoft and Meta approach privacy.
Login
