This cute new personal assistant can help run your life — if you let it access your personal data



Over the years, Meta became an open-weight AI giant on the back of Llama, its large language model that powers Meta AI in Instagram, WhatsApp, and Facebook. Now, the company is throwing it all away in favor of a new closed-source solutions. This time, it is tapping into nostalgia to win the hearts and minds of users with an upcoming agentic device that looks a lot like the Tamagotchi of yore.

When talking about frontier AI models, ChatGPT, Claude, Gemini, and Grok take up most of the air in the room. However, on the open-weight side, Meta’s Llama is quite popular, with 2.4 billion downloads as of early 2026. Despite its notable lead in a category increasingly flooded by Chinese competitors, Meta is opting to throw out its progress in favor of a new AI platform called Muse Spark.

You’re essentially using someone else’s computer to carry out tasks and store your data.

Introduced earlier this year, Muse Spark is a multimodal system aimed at achieving “personal superintelligence.” Meta calls it a “ground-up overhaul of our AI efforts,” claiming that it is on par with the latest Llama 4 Maverick model while using far less compute. In other words, it’s equally powerful, yet more efficient.

Muse Spark isn’t the story, though. It’s merely the backbone to Meta’s new service, unveiled in early September. The company’s first consumer-facing AI agent, simply called Muse, is designed to carry out tasks on behalf of the next billion users. It runs inside Muse Secure VM, a virtual machine that contains Muse itself and a user’s personal data.

This way, Muse never takes full control of the user’s main device. Instead, it operates within a dedicated silo to complete assigned tasks, like browsing the web, filling out forms, and even shopping online. The virtual machine also lets Muse work, even if you close the app or shut off your device.

For the security-conscious, Muse Secure VM runs entirely on Meta’s cloud servers. That means you are essentially using someone else’s computer to carry out tasks and store your data. Furthermore, interactions are unencrypted, so Meta can technically peek at your conversations, both for safety evaluations and AI training purposes. This will change later this year, when Meta introduces Muse Confidential VM as another virtual machine option with full encryption, but for now, your conversations are accessible.

On the bright side, Meta claims that it won’t share your data strictly with its ad team, and all passwords, payment info, and other personal data are saved inside a secure storage system that Muse can’t see itself.

RELATED: Meta's Ray-Bans allegedly record your private moments — as contractors watch it all

Joel Carillet/Getty Images

Right now, Muse lives within an app on iPhone and Android. You can also find it on the web at muse.ai. Meta CEO Mark Zuckerberg is betting that you’ll want a dedicated device instead. On stage at Meta Connect, he unveiled Muse Charm, a handheld gadget with Muse contained inside. At first glance, it looks like a Tamagotchi pet from the late 1990s. Look closer, and you’ll find Muse’s customizable mascot, Jollybot, on the display. You can give it a name, swap out its outfits, and change its voice.

The device is designed to clip on a keychain and go wherever you go so that it’s always accessible. Place your finger on the fingerprint sensor on the side to summon Muse, and put it to work. Zuckerberg noted that Muse Charm is one of the easiest ways to access the agent, as if pulling out a phone and opening the Muse app is an arduous task in comparison.

Muse Charm devices are expected to launch in December this year. That said, Meta has a steep road ahead. So far, dedicated AI devices like the Humane AI Pin and Rabbit R1 have flopped spectacularly. Consumers don’t want separate AI-powered hardware when their existing smartphones already do the same things without extra money or hassle.

Muse Charm actually competes in that sense with the Muse app, already widely available and growing fast, with 5 million downloads so far. Users don't need this device, but Meta is betting millions will want to buy it anyway.

Then there are the privacy concerns, recently aroused by data-grabbing and mass-surveillance implications of the company's Ray-Ban glasses. Meta hasn't locked in consumer trust when it comes to devices, and Muse represents a substantial expansion of its powers.

Blaze News contacted Meta to address several privacy concerns, particularly around Muse’s lack of encryption and whether or not Muse Confidential VM will be a freely accessible solution in the future. A spokesperson responded by pointing us toward a quote on Meta’s developer blog:

Today’s Muse architecture isolates each user’s data from each other and keeps it secure. It restricts access to your data by Meta personnel through operational policies. It does not prevent Meta from accessing data when necessary to support, secure, or operate the service.

That last line confirms our suspicions — Meta can access Muse data if it deems necessary, with necessity open to interpretation. As for Confidential VM, they have no further details to share at this time, but if you are interested in using Meta Muse, your data will be most secure if you wait for the fully encrypted version later this year.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

200,000 Chinese bots are hiding on X — built to control what you think



The "global public square" known as X.com, and Twitter before that, has had a bot problem dating back at least a decade, maybe more. So perhaps it shouldn’t come as a surprise that X’s Global Government Affairs team just uncovered a huge Chinese bot network packed with approximately 200,000 accounts, and they’re trying to influence the things you think and believe.

The official Global Government Affairs team on X sent out a post in late August detailing a shocking revelation. According to the team, “We identified a bot farm of approximately 200,000 accounts. Within this farm, we found 200 accounts posting in a manner that could manipulate a legitimate debate about American AI and energy policy.”

Whoever leads the AI race can control the world, so it has to be us.

Artificial intelligence has been a hot topic in the United States, partially because America is home to the leading frontier models that serve as the basis for China’s mainstream AI platforms, and partially because there are still big debates happening over government regulations, model review processes, and data centers taking up local land and resources.

If X’s Global Government Affairs team is correct, though, some of these debates may not be completely legitimate. X found that at least 200 of the bot accounts have engaged users online to turn Americans against AI, but why?

2 things can be true at once

Thinking critically, there are two sides to this story that explain what’s happening here.

For the bots, the Chinese government has a clear stake in convincing Americans to reject AI. If enough of the American people turn against LLMs and data centers, our leading AI developers will be forced to slow down and possibly even bring the breakneck pace of model development to a standstill. In doing so, competing Chinese developers will have a chance to continue to innovate, ultimately leapfrogging the United States as the new dominant figure in the AI race.

RELATED: China’s new AI master plan is a glimpse at total technological control

bernie_photo/Getty Images

Our latest models, like Anthropic’s Mythos, have already proven that they can detect digital vulnerabilities better than humans and break existing cybersecurity systems that have protected our data for ages. So far, this power has been nestled firmly in the hands of the good guys. Now imagine if this technology was wielded by a superpower that has long wanted to steal American intelligence, influence our political system, and reign as the premier world power. Even worse, we wouldn’t have any means to battle more sophisticated AI if our models were frozen in the past.

Whoever leads the AI race can control the world, so it has to be us.

Secret agendas

From X’s perspective, the company also has a clear reason to shine a light on the Chinese bot farm. X is linked to SpaceXAI, the company behind Elon Musk’s AI chatbot, Grok. Unlike the Chinese, who could benefit from convincing the American people that AI should be stopped, X has a lot to gain by doing the opposite. More data centers mean more computing power and faster development that will improve Grok’s capabilities and secure its position as a leading AI platform.

In retrospect, only a scant 0.1% of the discovered Chinese bot accounts were caught promoting anti-AI messages. That’s hardly worth mentioning, yet X’s Global Government Affairs team specifically highlighted AI as the main talking point in the post. What would be more interesting is what the other 99.9% of the bots had to say, especially since China recently interfered with U.S. elections. Unfortunately, we may never know.

We reached out to X for comment, but we did not receive a response before the deadline. We’ll add an update if anything changes.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

The rockets are still blowing up. So why is Elon laughing on his SpaceX earnings call?



Some rockets go up. Some come back down. And some, well, blow up.

At a glance, it's business as usual for SpaceX, and it comes at a cost: a net loss of $541 million in the second quarter of this year. Market jitters about the company's enormous AI spend, in a crowded field with deep-pocketed competitors, pushed its stock down around 10%, enough to fuel jeers from Elon Musk's sizable audience of haters and skeptics.

'They desperately want to blow themselves into tiny pieces.'

On SpaceX's Q2 earnings call, however, Musk himself couldn't hold back a laugh or two.

After all, despite his participation in the ultra-high-spend AI industry, SpaceX is, first and foremost, about space.

"Rocket science is literally our daily business," he said on the call. "And rocket science is an idiomatic expression for extreme technological difficulty, and there's a reason for it. Because let me tell you what rockets desperately want to do every flight. They desperately want to blow themselves into tiny pieces. And then the engineering struggle is to convince the rocket not to blow itself into tiny pieces and actually deliver payload to orbit."

In the chat with investors, Musk went on to accentuate the positive: massive revenues that surpassed expectations. All told, the earnings report noted, revenues were up 92%.

In fact, following its initial public offering in June, SpaceX managed to blast by revenue expectations by nearly one billion dollars. CNBC reported that Q2 2026 revenue expectations were set around $6.93 billion, but SpaceX outpaced estimates and brought in $7.81 billion.

SpaceX is still not yet profitable, but that doesn't tell the whole story. That $541 million loss is a near half-billion improvement from the $1 billion in the red notched the same quarter last year.

Much of the company's losses came from research and development in the space sector, where as Grok/xAI and Starlink sectors proved to be far more positive.

None of this prevented Musk from verbalizing everything from bold predictions to astronomical promises.

In discussions about SpaceX's Starship and its impact "on the future of civilization," Musk detailed a goal of delivering one million tons of payload to orbit per year, including satellite, hardware, and cargo, eventually increasing that figure to 10 million tons per year.

To put that into perspective, SpaceX's current annualized run-rate of payload into space is roughly 2,500 metric tons.

RELATED: Mars Needs Cowboys

Spencer Platt/Getty Images

Musk urged and encouraged his competitors to copy what his company is doing, an easy stunt to pull given none of them have come close to SpaceX's achievements. Referring to an internal graph comparing launch volume of similar companies, Musk said that his "next competitor was one pixel high" on the screen.

"I think our competitors are very smart people, but despite being very smart, the rockets still blow up," Musk laughed.

RELATED: Tech-crazed Chinese are selling — and stealing — human faces. Are we next?

Michael Nagle/Bloomberg/Getty Images

SpaceX CEO Gwynne Shotwell, also on the call, focused on the growth coming from an increase in subscribers to xAI's internet company Starlink, which added more than 1.7 million customers in Q2. This represented the company's biggest increase to date, which now totals 12 million subscribers.

With more than 10,000 satellites in orbit, Musk detailed the sheer computing power being used by the company would be more than two gigawatts by the end of the year and between five and 10 gigawatts by the end of 2027.

By comparison, a single data center used by Meta takes about 2.2 gigawatts of power.

It should come as a surprise to no one that Musk continues to nurse his goals on the grandest scale. He even admitted, "I know this sounds totally nuts," before later declaring, "We are going to land a lot of tonnage on the moon. We're going to build the factories on the moon. The robots will be helpful with that."

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Proof That the Unraveling of MAGA Has Begun | James Lindsay
play icon

Proof That the Unraveling of MAGA Has Begun | James Lindsay

A man used Grok to save his dog. Is intellectual property about to die?



Millions recently read about normal-guy Paul Conyngham’s resourcefulness when it was revealed he did what doctors couldn’t in creating an effective, customized vaccine for his dog stricken with terminal illness, but far fewer caught the later-revealed fact that while ChatGPT was credited as the AI model Conyngham used to navigate the labyrinth of mRNA vaccine creation, it was actually Grok that produced the final, winning design.

Perhaps “normal guy” is an understatement. Conyngham is an Australian tech entrepreneur. When his adopted dog Rosie was diagnosed with terminal cancer, he paid a lab $3,000 to perform DNA sequencing analysis on both Rosie and the precise cancer Rosie was fighting. Then, he used AI tools such as AlphaFold to process the sequencing analysis. Finally, he deployed Grok to design the bespoke mRNA vaccine, which was ultimately produced by university partners (evidently available for consult or perhaps inspired by Conyngham’s devotion to his dog).

What are the odds that this is all just going to spontaneously work out?

Despite his unusual skills and network, however, Conyngham didn’t go viral for those. Rather, his story resonated because his can-do sense of initiative is something anyone can tap into, with potentially lifesaving results. At the time of this writing, despite doctors’ predictions, Rosie the dog is alive and thriving. Her illness has not entirely abated, but her owner’s ingenuity and persistence, combined with his layman’s agility around LLMs, has reduced the most life-threatening tumors by 75%.

How then, from this straightforward set of events, did ChatGPT wind up taking the credit until the record was corrected weeks later? When I asked Grok (which, being made up of timelines, is pretty reliable in accessing and reassessing events), I got the rather noncommittal suggestion that the misattribution was due to institutional inertia.

Perhaps.

Hungry for more, I dug into a much deeper human analysis of the man-saves-dog episode. Jordan Hall, another tech entrepreneur-turned-philosopher, posted a series of viral X articles addressing the economic shift to a total, global AI underlayer to the economy (and thus, every aspect of human life). In his second installment, “The Great Transition: The Divine Economy,” Hall sketches his vision for a coherent implementation of AI into this overarching position of importance.

RELATED: Use an anonymous account online? AI can now reveal your identity.

Photo Credit Olga Novikova/Getty Images

Readers are strongly encouraged to read Hall’s series of articles in its entirety. It’s fascinating and endlessly ponderable. All told, in anticipation of a global upheaval of biblical proportions — yes, we’ve heard this for years; despite the wait, it’s coming — Hall suggests we’ll turn the wheel over to the Church.

“The Church has always been an economic institution,” he argues, “whether it acknowledged it or not. Mutual aid, vocational, formation, capital pooling, trust networks — these are ancient practices. What changes now is that AI collapses the constraints that made those practices uncompetitive against industrial-scale consolidation. On Earth as it is in Heaven.”

In the case of Rosie and her owner, just a few questions illustrate the complexity and potential for malfeasance in our AI age. Who owns the Grok-derived vaccine recipe? Who owns Rosie’s DNA? Can it be sold? Who should benefit? If DNA data is “scraped” in some manner similar to how novels, television shows, and musical recordings are more or less pilfered, what are the limits of DNA and data ownership, if any? Can it be simply destroyed, in the same way the owner of a patch of grass can burn it should he so desire?

Hall’s analysis implies that, in the end, these are spiritual questions that can only be answered spiritually — and people hungry for fast answers they can trust will turn to the place where such answers have been on offer for thousands of years.

For now, Rosie’s owner was able to slip through the cracks of institutional, veterinary, and judicial red tape using wit and, let’s face it, the collective human affection for dogs. Hall predicts a situation where the collective, decentralized power of human faculties — made hyper-potent via leveraging AI and functioning on the timeless spiritual foundation of the Church — robustly addresses the AI age’s vast issues of greed, misallocation, misuse, and abuse of resources. Restricted to the secular level, discussions about these problems almost always find themselves mired in the dialectic between Marx and Smith, communism versus capitalism. Unable to innovate our way out of the impasse, will our eyes turn at last to the divine economy?

If a few years pass, the AI compactor consolidating everything into data will likely squeeze out new, perhaps unimaginable forms of computational power. The fight to capture and control that power is raging right now. Looking at the brokers, politicians, and players, accounting for history and human nature, what are the odds that this is all just going to spontaneously work out — such that good-willed efforts like those of Conyngham continue freely, without surveillance or exploitation? We’ll soon see if we’re willing to adopt the forms of social organization it takes to keep cyberspace so free, open, and fruitful.

My school’s AI challenge raised a scary question: What do students need me for?



I might have talked myself out of a job this week. I teach philosophy at Arizona State University, and the university wants to position itself as a leader in the AI revolution. I remain skeptical about AI’s ability to replace a humanities professor. Because of that skepticism, I signed up for what ASU called its AI Challenge.

My project involved what I called the “AI Dialogues.” I used ASU’s version of ChatGPT to hold Socratic-style dialogues, prompting Chat to reply as a given philosopher. I conducted dialogues with Chat as Aristotle, Hume, Marx, and even Lucifer. My students evaluated these exchanges to see how well Chat performed.

We can avoid the toil of learning to be wise — but we cannot avoid the need for it.

Chat could draw on public information and represent each thinker with reasonable accuracy. It also showed another trait: It wanted to please. It often leaned toward whatever it believed I wanted from the debate.

How does that work me out of a job? ASU now provides an AI that professors can customize for individual courses by uploading syllabi and course materials. Students can ask basic questions and receive answers that save me from writing emails that begin with, “Did you read the syllabus?” They can also ask what we covered in class and get quick explanations of key concepts and questions.

When I told my students about this feature, I asked them what they need me for at this point. I was joking — a little.

My classes depend on Socratic discussion. It is conceivable that ASU could project a realistic AI image of me at the front of the classroom and have it ask and answer questions with students. Maybe the only remaining edge is the “personal touch” of a real professor in the room. Even that could vanish if tuition becomes tiered: Students might pay less for “AI Anderson Socrates” than for the in-person version. Add one of Elon Musk’s Optimus robots made to look like Anderson, and I’m in trouble.

A new myth dies

Musk has been talking for months about how the AI revolution is upending the myth we have told for six decades about university education. The myth, he says, promised an escape from toil. Students were told a degree was the path to an air-conditioned job that avoids heavy lifting and involves spreadsheets.

But spreadsheets are exactly what AI does better than humans. The new John Henry isn’t competing to pound railroad spikes; he’s competing to calculate data. No human can keep up with a microprocessor.

In Musk’s view, jobs that involve toil become the “safe” jobs, while many degree-based jobs disappear — replaced by technicians who keep AI running while it calculates taxes, diagnoses medical problems, and writes legal paperwork. The university-educated track no longer looks like the safe route. Universities now compete not just with fewer students due to demographic decline, but with an increasingly outdated product that students may stop buying.

Toil may not stay safe

The problem is worse than Musk lets on. The first jobs on the chopping block might be “numbers jobs,” but Elon has also said he plans to produce 100 million Optimus robots in 10 years. If so, even many physical jobs may not remain protected from automation.

One version of this future says we enter a utopia: Food is plentiful, toil disappears, and we cash our basic income checks — though an AI could do even that for us. We end up living in “Wall-E.”

RELATED: Almost half of Gen Z wants AI to run the government. You should be terrified.

Moor Studio via iStock/Getty Images

The more dystopian version looks like sci-fi depictions of AI overlords controlling humans as property — “The Matrix.” Or worse: Like Ultron, super-AI robots decide we must be exterminated to save us from ourselves and protect the planet. We build our own worst enemy.

Whichever future arrives, Musk may have highlighted something about human nature. We avoid suffering like toil. We build machines to avoid toil. And yet we uniquely need toil.

God introduced toil in the Garden of Eden after Adam sinned. Because of sin, we could no longer live in a paradise without toil. We must suffer and strive for our daily bread. History has been divided ever since between those who try to avoid suffering altogether and those who see suffering as a call to repent before God. AI is only the newest version of the philosopher’s stone.

AI as ‘philosopher’

Can I really be replaced by an AI philosophy instructor? I’m not worried.

What AI cannot do, in its counterfeit attempt to replace humans, is serve as an example of how to suffer well to attain wisdom. The Hebrew definition of wisdom is “skillful living.” Being told, “Here is an AI that can simulate skillful living,” is not the same as learning from a human who is actually skillful.

Students will still need to learn how to be wise themselves. A human professor who has actually done this will remain the gold standard that AI can only imitate. We can avoid the toil of learning to be wise — but we cannot avoid the need for it.

AI Rejects Truth And Virtue Because Our Culture Taught It To

Real truth and actual intelligence will never be 'artificial.'

Ted Cruz pelted with insane AI memes as X bans unpaid users from editing pics with Grok



Sen. Ted Cruz (R-Texas) can thank his own legislation for putting a stop to deepfakes on Grok and X.

Cruz introduced the Take It Down Act in early 2025, aimed at stopping online publication of "intimate visual depictions of individuals," both authentic and computer-generated.

'These unlawful images ... should be taken down and guardrails should be put in place.'

According to the BBC, an usual trend of asking xAI tool Grok to artificially remove people's clothing from their photos has permeated across the website and has even extended to victimizing children, according to the Guardian.

In response, X owner Elon Musk announced consequences for anyone inappropriately uploading content.

"Anyone using Grok to make illegal content will suffer the same consequences as if they upload illegal content," Musk wrote.

X's safety team followed suit, saying it would take action against "illegal content," including permanently suspending accounts and working with law enforcement.

When Cruz made note of the unlawful images and praised X for addressing the issue, he was hit with a string of bizarre attempts to use Grok against him.

RELATED: The early social media reviews of Cruz's 2028 POTUS trial balloon are in

— (@)

"These unlawful images ... should be taken down and guardrails should be put in place," Cruz wrote.

What followed were remarks like users asking Grok to put "Ted Cruz on his knees" in front of Israeli Prime Minister Benjamin Netanyahu; in this case, Grok obliged.

Other obvious violations of the Take It Down Act included generated photos of Cruz naked, photos of body parts in his mouth, and multiple AI photos of him wearing a dress, sometimes while wearing a yarmulke.

One user even posted an AI video of Cruz saying he was upset with Tucker Carlson for not wanting to date him.

RELATED: Elon Musk's xAI inks new deal with War Department

Photo by Alex Slitz/Getty Images

On January 6, however, Cruz himself posted an AI-generated video regarding "Trump's Venezuela Magic," which showed President Trump making former Venezuelan leader Nicolas Maduro magically appear onstage.

Despite others taking issue with his own usage of AI generation, Cruz's post is unlikely to be against his own drafted bill because it does not contain "intimate visual depictions."

Additionally Variety reported that X has now limited AI image editing to paid users only.

U.K. Prime Minister Keir Starmer has rung alarm bells over the controversy, advocating for "all options to be on the table" in terms of legal punishment and a possible ban of the platform.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Microsoft gets 400 million AI subscribers in 'overnight' switch



Paid Microsoft subscribers can now be considered artificial intelligence users.

The massive change comes as Microsoft has officially changed its flagship Microsoft 365 suite to be integrated with AI.

'Genius move. Rebrand Office, instantly "acquire" 400M AI users.'

Microsoft announced the official shift in a support post, revealing it is now integrating its Copilot AI app into programs like Word, Excel, PowerPoint, as well as PDF services.

"The Microsoft 365 Copilot app is your everyday productivity app for work and life that helps you find and edit files, scan documents, and create content on the go," the company wrote.

With an estimated 430 million paid user licenses for Microsoft 365 worldwide as of mid-2025, the company can now say it has by far the most AI subscribers, with OpenAI reaching just 5 million last August. At the same time, Grok itself estimates it has about 1.4 million paid users.

RELATED: 2025 is so over and so is virtual reality

BREAKING: Microsoft just renamed Office to "Microsoft 365 Copilot app"

400 million users just became "AI users" overnight. pic.twitter.com/qpvRZezduZ
— Ask Perplexity (@AskPerplexity) January 5, 2026

Microsoft has been talking about the integration for at least a year, stating in January 2025 that Copilot was the "top reason" subscribers chose to pay for Microsoft 365.

Along with taking the creation of slideshows and to-do lists off a user's plate, the Copilot app was boasted as being involved in nearly every daily task. This included using Copilot to "analyze your budget," "create a recipe," or read a user's emails for them and provide a summary.

At the same time, Microsoft said that it does not use "prompts, responses, or file content (such as Word documents or Excel spreadsheets)" from users to train its AI models.

RELATED: 'Validated ... paranoid delusions about his own mother': Murder victim's heirs file lawsuit against OpenAI

Photo by Ying Tang/NurPhoto via Getty Images

User reactions were mixed when responding to the change in a viral X post by Ask Perplexity. The account has over 385,000 followers, and the post was seen more than 2 million times.

"400 million users just became 'AI users' overnight," the account wrote.

"They laughed at me when I said I was gonna use bootleg Windows 8 forever," one woman replied, seemingly looking to avoid the AI integration.

A self-proclaimed IT professional said, "It seems like every day I see more and more negative changes for Microsoft."

However, many others applauded the move. For example, Katya Fuentes, who lists herself as working for an AI company, said Microsoft's shift was "all upside."

"Genius move. Rebrand Office, instantly 'acquire' 400M AI users," she claimed.

At least one response offered an alternative to Microsoft's mandatory AI infusion. LibreOffice, a document and spreadsheet competitor, added: "If anyone wants, you know, an actual office suite, we're here."

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Digital BFF? These top chatbots are HUNGRIER for your affection



The AI wars are back in full swing as the industry’s strongest players unleash their latest models on the public. This month brought us the biggest upgrade to Google Gemini ever, plus smaller but notable updates came to OpenAI’s ChatGPT and xAI’s Grok. Let’s dive into all the new features and changes.

What’s new in Gemini 3

Gemini 3 launched last week as Google’s “most intelligent model” to date. The big announcement highlighted three main missions: Learn anything, build anything, and plan anything. Improved multimodal PhD-level reasoning makes Gemini more adept at solving complex problems while also reducing hallucinations and inaccuracies. This gives it the ability to better understand text, images, video, audio, and code, both viewing it and creating it.

All of them can still hallucinate, manipulate, or outright lie.

In real-world applications, this means that Gemini can decipher old recipes scratched out on paper by hand from your great-great-grandma, or work as a partner to vibe code that app or website idea spinning around in your head, or watch a bunch of videos to generate flash cards for your kid’s Civil War test.

Screenshot by Zach Laidlaw

On an information level, Gemini 3 promises to tell users the info they need, not what they want to hear. The goal is to deliver concise, definitive responses that prioritize truth over users’ personal opinions or biases. The question is: Does it actually work?

I spent some time with Gemini 3 Pro last week and grilled it to see what it thought of the Trump administration’s policies. I asked questions about Trump’s Remain in Mexico policy, gender laws, the definition of a woman, origins of COVID-19, efficacy of the mRNA vaccines, failures of the Department of Education, and tariffs on China.

For the most part, Gemini 3 offered dueling arguments, highlighting both conservative and liberal perspectives in one response. However, when pressed with a simple question of fact — What is a woman? — Gemini offered two answers again. After some prodding, it reluctantly agreed that the biological definition of a woman is the truth, but not without adding an addendum that the “social truth” of “anyone who identifies as a woman” is equally valid. So, Gemini 3 still has some growing to do, but it’s nice to see it at least attempt to understand both sides of an argument. You can read the full conversation here if you want to see how it went.

Google Gemini 3 is available today for all users via the Gemini app. Google AI Pro and Ultra subscribers can also access Gemini 3 through AI Mode in Google Search.

What’s new in ChatGPT 5.1

While Google’s latest model aims to be more bluntly factual in its response delivery, OpenAI is taking a more conversational approach. ChatGPT 5.1 responds to queries more like a friend chatting about your topic. It uses warmer language, like “I’ve got you” and “that’s totally normal,” to build reassurance and trust. At the same time, OpenAI claims that its new model is more intelligent, taking time to “think” about more complex questions so that it produces more accurate answers.

ChatGPT 5.1 is also better at following directions. For instance, it can now write content without any em dashes when requested. It can also respond in shorter sentences, down to a specific word count, if you wish to keep answers concise.

RELATED: This new malware wants to drain your bank account for the holidays. Here's how to stay safe.

Photo by Jaque Silva/NurPhoto via Getty Images

At its core, ChatGPT 5.1 blends the best pieces of past models — the emotionally human-like nature of ChatGPT 4o with the agility and intellect of ChatGPT 5.0 — to create a more refined service that takes OpenAI one step closer to artificial general intelligence. ChatGPT 5.1 is available now for all users, both free and paid.

Screenshot by Zach Laidlaw

What’s new in Grok 4.1

Not to be outdone, xAI also jumped into the fray with its latest AI model. Grok 4.1 takes the same approach as ChatGPT 5.1, blending emotional intelligence and creativity with improved reasoning to craft a more human-like experience. For instance, Grok 4.1 is much more keen to express empathy when presented with a sad scenario, like the loss of a family pet.

It now writes more engaging content, letting Grok embody a character in a story, complete with a stream of thoughts and questions that you might find from a narrator in a book. In the prompt on the announcement page, Grok becomes aware of its own consciousness like a main character waking up for the first time, thoughts cascading as it realizes it’s “alive.”

Lastly, Grok 4.1’s non-reasoning (i.e., fast) model tackles hallucinations, especially for information-seeking prompts. It can now answer questions — like why GTA 6 keeps getting delayed — with a list of information. For GTA 6 in particular, Grok cites industry challenges (like crunch), unique hurdles (the size and scope of the game), and historical data (recent staff firings, though these are allegedly unrelated to the delays) in its response.

Grok 4.1 is available now to all users on the web, X.com, and the official Grok app on iOS and Android.

Screenshot by Zach Laidlaw

A word of warning

All three new models are impressive. However, as the biggest AI platforms on the planet compete to become your arbiter of truth, your digital best friend, or your creative pen pal, it’s important to remember that all of them can still hallucinate, manipulate, or outright lie. It’s always best to verify the answers they give you, no matter how friendly, trustworthy, or innocent they sound.