IT General

Is your TikTok FYP inexplicably filling up with short clips of AI baby videos? If so, you're not alone, the AI baby videos are going viral across TikTok, Facebook, and Instagram.

These short AI-generated videos feature a cherubic baby-faced version of well-known figures, from world leaders like French President Emmanuel Macron to the stars of Friends. See also: Vin Diesel in Fast & Furious, Michael Jackson in "Thriller," and Judge Judy.

So, are these videos cute, creepy, or cringe? Cringe is in the eye of the beholder, and your take on this trend will be informed by your general fondness for both babies and AI in general. Regardless, the viral videos are yet more proof that the AI slop machine is working overtime.

"Slop" is the term for low-quality, AI-generated images and videos, and they can often go surprisingly viral. Some of the people sharing the clips may not even realize the videos are AI-generated, though we hope that goes without saying when you come across a video of baby Ice Cube (Lil' Ice Cube?) rapping "It Was a Good Day."

Where are all of these AI baby videos coming from? And how are they made?

The first question is easy to answer. Some videos are surely created by fans and AI enthusiasts, while others are churned out by content farms. Like those men shoveling coal into the Titanic's engines, the algorithms must be fueled, and content farms shovel a constant stream of memes, AI content, and stolen videos onto social media apps.

This particular trend started a couple of weeks ago, with AI enthusiasts making AI videos featuring talking babies on podcasts. The trend evolved to AI videos of celebrities as babies, which is where we are now. As mentioned earlier, the trend has even reached the highest level of world government.

How to make AI baby videos

Most creators are using a mix of different AI tools to create these baby videos, and you can find tutorials on YouTube and TikTok. However, I know what you're really wondering: Can I create these AI baby videos for free, and without being a tech wiz? Not really.

I tried creating an AI video using TikTok's new free AI Alive tool, which turns images into animated videos. The results were, ah, not great!

Credit: Timothy Beck Werth / Mashable Credit: Timothy Beck Werth / Mashable

To create an AI baby video from start to finish, you will need paid subscriptions to some popular AI tools. Some of these steps can be done with the free versions of the tools mentioned below, but generally, you need a paid subscription to generate a video.

1. Upload a picture of yourself to ChatGPT (or your AI tool of choice); use a prompt like this: "Create a baby photo based on this image."

2. Write a short script to go along with your video.

3. Use the ElevenLabs text-to-speech tool (or a similar tool) to create an audio file for your "baby."

4. Use an AI video tool like Hedra or Dreamina to generate a video based on your photo and audio file.

Of course, all this comes at a cost, and not just for AI subscriptions. According to a new report from MIT Technology Review, generating one 5-second AI video is the energy equivalent of running a microwave for a full hour.

Which AI video tool should you use? Google's brand new Veo 3 video generator is currently the most advanced video generator, but it also costs $250 a month to access as part of Google's AI Ultra Subscription. I was able to create a fairly realistic baby video using Veo in Gemini, but that's not exactly the point of the meme.

Most creators seem to be using Hedra. Per Tech Crunch, Hedra just raised $32 million, and its Character-3 model is powering a lot of these AI baby videos.

When Microsoft releases a new major update to Windows 11, the company wants you to update your computer, and therefore quickly deprecates the previous versions. In a relatively rare move, Microsoft is rolling out a few updates to Windows 11 23H2 and 22H2.

Thanks to cheaper CG and the rise of long, serialized TV shows from on-demand streaming services, it feels like we're in the golden age of book adaptations. Yes, not every show will be a hit or a good adaptation (looking at you, Wheel of Time), but so many books were unfilmable til recently.

More than 184 million passwords may have been compromised in a massive data leak affecting everything from social media logins to bank accounts, according to a new report published by cybersecurity researcher Jeremiah Fowler. Ironically, the database containing the leaked passwords was itself non-password-protected, according to Website Planet.

Fowler says he uncovered the publicly accessible and unencrypted database, which contains 184,162,718 unique logins and passwords to popular websites and apps. As spotted by our colleagues at ZDNet, the exposed logins and passwords are reportedly tied to email providers like Google, a range of Microsoft products, as well as social media platforms like Facebook, Instagram, and Snapchat. Credentials for other applications and services, such as Roblox, were also found in the database. Furthermore, according to Fowler, sensitive information from bank accounts, health services, and even government portals was stored in the database, which is no longer online.

Because Mashable has not been able to review or independently confirm that the database contained leaked data, we reached out to some of the companies implicated in the report. A Snapchat representative said it has not found any evidence of a data breach or vulnerability on their platform. We'll update this article if we hear back from any other companies.

While it's not yet clear exactly how the sensitive data in the database may have been stolen, the fact that platforms like Snapchat have not found any vulnerabilities in their network makes sense considering Fowler's assessment of the situation.

According to Fowler, the database showed signs that it was compiled with data from an infostealer malware. Fowler describes infostealer as a type of "malicious software designed specifically to harvest sensitive information from an infected system." This means that the more than 184 million passwords were likely stolen directly from the affected users themselves.

Infostealer malware can scrape user data stored in web browsers, including autofill data and cookies. Data stored in emails, including drafts and documents, as well as messaging apps, can also be harvested by simply tricking the targeted user into downloading the malware.

These sorts of data breaches have become increasingly common in recent years. For example, Mashable previously reported on the RockYou2024 leak, which is regarded as the biggest password leak ever. In July of last year, a malicious user shared this compilation of nearly 10 billion credentials on hacker forums. These large data breaches provide bad actors with a dangerous tool to carry out automated brute force attacks.

Mashable will update this post with any new information about the leak.

Spend enough time on TikTok, and you might wonder why the hell everyone is asking, "Who the hell put the muffins the freezer?"

The answer to that question is both simple and a little complicated. The simple answer is that it's a silly trend on TikTok with that bit of audio and, typically speaking, a short video posting some kind of reveal. The complicated answer is that this all started back on Vine, the now-defunct, short-form video platform that predated TikTok.

SEE ALSO: Vine is dead. Let's keep it that way.

But first, back in the present day, the trend has been all over TikTok lately. The trend typically involves somebody lip-syncing the audio, "Who the hell put the muffins in the freezer?" with someone else in the reveal saying, "I did." Oftentimes, the trend is purposefully cringeworthy since the audio is, well, pretty cringeworthy.

Sometimes, there are even literal muffins in literal freezers.

It's become so popular that it's made its way to A-list celebs like Bella Hadid.

That's it... That's the whole trend. It's a throwback, really, to the Vine days. Vine's 6-second limit often pared down humor to somebody just saying something in a weird way. Remember "FR E SH A VOCA DO?" That's what's happening here: The sound is odd, and the videos are cringey, short, and easy to film.

It's not surprising, then, that this trend dates back to Vine and, more recently, the year 2020. As Distractify wrote five years ago, the TikTok account @beanstalk_0 uploaded what appears to be an old Vine of a kid shrieking, "What the fuck?" then asking, in that unmistakeable tone, "Who the hell put the muffins in the freezer?" That post went viral.

Then, musician Tiagz transformed it into a song, which also went viral.

This is something Tiagz was known for at the time, with Pitchfork dubbing him in 2020 a "cringey remixer" who was "the most hated producer on TikTok." So yes, you're reading that right, the 2025 trend is a retread from 2020. Frankly, I didn't even remember it because, well, 2020 was a weird year.

So we've spent all this time questioning who the hell put the muffins in the freezer, when really we should be asking how long they've been there.

3D printers are now affordable and more user-friendly than ever before, but that doesn't mean there aren't some pitfalls still out there for people new to the hobby—or those who want to do serious business!

Have you been wanting to try out Linux, but aren't sure how? Since ChromeOS ships with a fully functional Linux shell, a Chromebook Plus might be the best choice for giving Linux a try without fully committing to a high-end Linux laptop from the start.

When people look at gaming PCs, they tend to imagine gigantic, honking monsters, but a gaming PC can actually get pretty small while still playing games. Case in point—this ASUS tiny PC packs some serious hardware capable of gaming.

Pokémon is one of the longest-running and most storied franchises in history. What started as a simple role-playing game for the Game Boy has evolved into a global multimedia empire spanning animated series, feature films, the Pokémon Trading Card Game, and more than 100 video games. With such a massive catalog, it's no surprise that a few oddballs slipped through the cracks. And unless you're a diehard enthusiast, chances are there are some titles you've never heard of.

Over the past few weeks, we’ve covered the rise of phishing scams sweeping across the US. Now, a new variant is making the rounds: scammy SMS messages targeting drivers under the guise of official DMV notices. These DMV text and ticket scams seem to be all over the country.

In recent weeks, Departments of Motor Vehicles in states like California, Florida, and New York have issued public warnings about fraudulent texts claiming to represent their offices. The messages allege that drivers owe unpaid tolls or have outstanding violations, and if they don’t pay up, they’re threatened with credit damage or even prosecution. Google Trends shows that interest is spiking in other states, too, like Utah and Illinois, suggesting the scam is rapidly gaining traction.

A more troubling detail is the misuse of state-specific legal codes to increase credibility. Most of the examples seen online show the same code: "15C-16.003," which is a real statute in Florida relating to car insurance, but is misused in messages sent to New Yorkers, where no such code exists.

This Tweet is currently unavailable. It might be loading or has been removed.

State agencies do not use text messages to deliver legal notices or collect fines. The recent DMV scams seem quite similar to the E-ZPass scam texts that have proliferated in the last few months. Scammers rely on exploiting confusion and targeting individuals who are less familiar with digital threats, often the elderly or those who aren’t tech-savvy. That's why the messages are sometimes filled with shoddy grammar and easily verifiable mistakes — scammers don't want to waste time on people who know better. The recent DMV ticket scams also use phrases like "DMV Final Notice," which increases urgency and makes people act before they had time to think clearly.

The most effective defense is simple: ignore the messages and do not click the links. You can also delete the message and report it as junk. If there’s any concern about outstanding DMV fees or violations, official state DMV websites allow users to securely check your status and verify if you have a ticket or any outstanding charges.

Have a story to share about a scam or security breach that impacted you? Tell us about it. Email submissions@mashable.com with the subject line "Safety Net" or use this form. Someone from Mashable will get in touch.

Updated on May 28 at 6:55 p.m. ET: On Wednesday, May 28, Bloomberg reported that Tesla will now be launching its robotaxi service in Austin on June 12. According to that report, the June 12 date "could still change" in the days ahead. You can read our original story on the Tesla robotaxi rollout below.

Elon Musk has said that he's now going all-in with his companies after shifting focus from his role as a special government employee for the Trump administration. And there's no shortage of problems for Musk to attend to, including Tesla's recent abysmal quarterly report and crashing Tesla sales numbers in Europe. Now, another big Tesla project may be in jeopardy.

Tesla is set to launch its long-awaited robotaxi program in Austin, Texas next week. However, according to a new report from Fortune, the city of Austin is not ready for Tesla's robotaxis just yet.

A small fleet of Tesla robotaxis is already up and running in Austin and San Francisco, serving an "early set of employees" in the two cities as part of an initial testing phase. Next week's broader rollout would bring about 10-20 Tesla Model Y self-driving cars to Austin's streets, serving the public for the first time in a robotaxi capacity. If successful, Tesla plans to scale the number of robotaxis into the thousands.

Fortune reports that Tesla held initial meetings with both state and city agencies in Austin a few weeks ago. However, according to that report, officials with Austin’s transportation department, the city's emergency first responders, and federal regulators say that Tesla has failed to deliver crucial information regarding the service, which is supposed to go live in just a few days.

Per Fortune, the company has so far failed to share first responder plans or guides with Austin's transportation department and fire department. That information could be critical if a robotaxi crashes and first responders need to rescue passengers, put out a vehicle fire, or use "jaws of life" rescue tools. In fact, Austin officials told Fortune Tesla hasn't even told the city what level of autonomy the Tesla robotaxis will have at launch, meaning Austin isn't sure if Tesla's service will require human supervision or not.

While other self-driving car companies, like Waymo, are currently operating in Austin, Tesla uses its own technology. Waymo's vehicles have large sensors sitting atop the cars in order to navigate the roads, while Tesla's vehicles utilize video cameras and AI.

Tesla has been in touch with state and city agencies and has assured them that more information is on the way, according to officials the Fortune reporter spoke with. The company has been present in Austin for at least a few weeks now in order to run some initial testing. However, Fortune described the preparations so far as rushed.

Musk has been hyping up Tesla's robotaxi capabilities for years, even as competitors beat him to launching their own self-driving ride services. Meanwhile, Musk is overseeing preparations for the latest Starship launch. So far, the Starship rockets have had an unfortunate tendency to explode.

Will Tesla's robotaxi service have similar problems after launch? We'll know more soon.

Congressional leaders are begging the federal government to fulfill its promises to modernize the national wireless emergency alert system, as severe weather events plague communities.

In a letter sent this week to Federal Communications Commission Chair Brendan Carr, California Representative Nanette Diaz Barragán called out the FCC for delaying a years-long effort to provide multilingual Wireless Emergency Alerts for wildfires, hurricanes, earthquakes, tsunamis, and 911 outages. The importance of such communications was made more clear during the January wildfire disaster in Los Angeles.

The plan was approved in 2023 and set to go into effect in January, providing 18 alert templates in 13 of the most commonly spoken languages in the U.S. for use by nationwide authorities and all major wireless providers. The system would bypass the need for translators, speed up wireless alerts across devices, and require mobile service providers pre-install multilingual alerts.

But the Trump administration's new leadership has balked at deploying the system.

SEE ALSO: Explicit deepfakes are now a federal crime. Enforcing that may be a major problem.

“You’ve got about 68 million Americans that use a language other than English and everybody should have the ability to to understand these emergency alerts. We shouldn’t be looking at any politicization of alerts — certainly not because someone’s an immigrant or they don’t know English," Barragán told the Los Angeles Times. Carr was previously an advocate for the alert modernization effort, but has delayed publishing the new rules in the Federal Register.

Instead, the FCC has spent the last four months investigating companies, organizations, and federal contractors for diversity-related spending and content decisions, including NPR and PBS broadcasts, Disney, and CBS. In March, Verizon capitulated to the administration's anti-DEI goals and ended its "DEI-related practices" in order to finalize a $20 billion purchase of broadband provider Frontier Communications.

Meanwhile, advocacy coalitions, including the AAPI Equity Alliance, called the stalled effort a "real dereliction of duty" and a direct attack on immigrants, the LA Times reported.

Emergency alerts have become a crucial lifesaving tool as natural disasters increase in size and frequency, a foreseen consequence of the global climate crisis. But the country's alert system has struggled to scale up in accuracy and access, and a lack of federal funding for organizations like the National Oceanic & Atmospheric Administration (NOAA) and the National Weather Service (NWS) are doing little to help.

Apple has bought its first video game studio, and will finally be publishing games. The acquired Vancouver-based studio, RAC7, is a small team founded in 2014 by Jason Ennis and Jesse Ringrose.

One of the things that make video games different from other media, is that you can often control what happens in the story. Sometimes the best games are ones that leave tough yet memorable choices with you.

It only took 16 years, but there's finally an official version of WhatsApp for the iPad. If you're tired of using WhatsApp's web version on your iPad and wish there were a better way, this is exactly what you've been waiting for.

Is your computer monitor no longer sleeping when it's supposed to? This is a problem I ran into and couldn't figure it out—until I found one simple command that revealed everything.

Lawyers representing Anthropic recently got busted for using a false attribution generated by Claude in an expert testimony.

But that's one of more than 20 court cases containing AI hallucinations in the past month alone, according to a new database created by French lawyer and data scientist Damien Charlotin. And those were just the ones that were caught in the act. In 2024, which was the first full year of tracking cases, Charlotin found 36 instances. That jumped up to 48 in 2025, and the year is only half way over. The database, which was created in early May, has 120 entries so far, going back to June 2023.

SEE ALSO: More concise chatbot responses tied to increase in hallucinations, study finds

A database of AI hallucinations in court cases shows the increasing prevalence of lawyers using AI to automate the grunt work of building a case. The second oldest entry in the database is the Mata v. Avianca case which made headlines in May, 2023 when law firm Levidow, Levidow & Oberman got caught citing fake cases generated by ChatGPT.

The database tracks instances where an AI chatbot hallucinated text, "typically fake citations, but also other types of arguments," according to the site. That means fake references to previous cases, usually as a way of establishing legal precedent. It doesn't account for the use generative AI in other aspects of legal documents. "The universe of cases with hallucinated content is therefore necessarily wider (and I think much wider)," said Charlotin in an email to Mashable, emphasis original.

"In general, I think it's simply that the legal field is a perfect breeding ground for AI-generated hallucinations: this is a field based on load of text and arguments, where generative AI stands to take a strong position; citations follow patterns, and LLMs love that," said Charlotin.

The widespread availability of generative AI has made it drastically easier to produce text, automating research and writing that could take hours or even days. But in a way, Charlotin said, erroneous or misinterpreted citations for the basis of a legal argument are nothing new. "Copying and pasting citations from past cases, up until the time a citation bears little relation to the original case, has long been a staple of the profession," he said.

The difference, Charlotin noted, is that those copied and pasted citations at least referred to real court decisions. The hallucinations introduced by generative AI refer to court cases that never existed.

Judges and opposing lawyers are always supposed to check citations for their own respective responsibilities in the case. But this now includes looking for AI hallucinations. The increase of hallucinations discovered in cases could be the increasing availability of LLMs, but also "increased awareness of the issue on the part of everyone involved," said Charlotin.

Ultimately, leaning on ChatGPT, Claude, or other chatbots to cite past legal precedents is proving consequential. The penalties for those caught filing documents with AI hallucinations include financial sanctions, formal warnings, and even dismissal of cases.

That said, Charlotin said the penalties have been "mild" so far and the courts have put "the onus on the parties to behave," since the responsibility of checking citations remains the same. "I feel like there is a bit of embarrassment from anyone involved."

Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

As you can probably tell by all the grad party invitations plastering your fridge, it's officially high school graduation season. Shopping for a high school grad can be daunting, though: Do you go for something fun? Something practical? What do these kids even like these days?! Don’t worry — we are here to help you impress your favorite high school graduate with a great gift. If they're planning on moving into a college dorm or to their own apartment, you know they'll need plenty of basic, essential items. The good news is that they're likely starting with a clean slate, so you have a ton of room to work with. If we're being honest, the gift most grads really want is cold hard cash. You also can’t go wrong with a gift card — but if you're looking to get more personal, we have lots of solid suggestions below that your high school grad will love.

TL;DR: Online safety for the whole family is easy with this lifetime AdGuard Family Plan, now just A$24 (reg. A$265) with code FAMPLAN through 1 June.

Opens in a new window Credit: AdGuard AdGuard Family Plan: Lifetime Subscription AU$24
AU$265 Save AU$241 Get Deal

The internet isn’t just cluttered with ads; there are also trackers and other hidden threats. AdGuard keeps you safe online by blocking intrusive ads, stopping trackers, and safeguarding your privacy across all your devices.

Right now, you can grab lifetime protection for the whole crew with the AdGuard Family Plan. Cover up to nine devices for only A$24 (reg. A$265) with code FAMPLAN now through 1 June.

Enjoy a safer and smoother browsing experience

The internet should be fun, not frustrating. AdGuard tackles the most annoying parts of online life, like pop-ups, autoplay videos, and banner ads, so that you can enjoy your time browsing, streaming, and scrolling again.

Aside from blocking ads, AdGuard also protects your privacy. Your information will stay safe from trackers and activity analyzers, and you'll be shielded from malware and phishing websites that you or your family members could fall victim to. 

Parents will love AdGuard's parental control feature, which keeps your kids safe from inappropriate content online.

This Family Plan lets you protect up to nine devices with AdGuard, so you can make sure everyone's devices stay safe. And it works with Android and iOS operating systems alike, so it can be used on tablets, smartphones, and laptops.

Secure your own AdGuard Family Plan for life for just A$24 with code FAMPLAN now through 1 June. 

StackSocial prices subject to change.

GET $110 OFF: As of May 27, the GoPro Hero11 Black Mini is on sale for $139.99, down from its standard price of $249.99, for a savings of $110, or 44% off.

Opens in a new window Credit: GoPro GoPro Hero11 Black Mini $139.99 at Best Buy
$249.99 Save $110 Get Deal

It's rare for a camera model to define a whole style of photos and video. Just as the Polaroid was once the camera of in-the-moment friends and family shots, the GoPro now embodies no holds barred adventure footage. Built to splash into the waves or survive hard bumps on mountain biking trails, these cameras are made for extreme footage.

Right now, you can get the GoPro Hero11 Black Mini for $110 off, just in time for summer.

SEE ALSO: Go electric with $800 off the Bluetti Solar Generator Elite 200 V2 The best tech deals

• Roku Ultra 4K Ultimate Streaming Player (2024 Release) — $79.99 (List Price $99.99)

• Beats Pill Bluetooth Speaker — $99.95 (List Price $149.95)

• Roborock Qrevo Master Robot Vacuum and Mop — $899.99 (List Price $1599.99)

• Peloton Bike — $1,145.00 (List Price $1445.00)

• Apple AirTag (4-Pack) — $74.99 (List Price $99.00)

The Hero11 is built to record steadily, even when the going is bumpy. Horizon Lock does the work that a gimbal or stabilization software would. The camera records in up to 5K with a 24.7 megapixel sensor and a 16:9 aspect ratio. It has a maximum speed of 240 Fps. The Mini records to MicroSD and connects wirelessly via Bluetooth.

The Hero11 Mini is even more portable and mountable than the Hero11. It is 0.29 pounds and just over two inches square. It has two sets of mounting fingers, for versatile mounting options from bottom and rear.

Right now, May 27th, get the GoPro Hero11 Black Mini for $139.99 at Best Buy and save $110, or 44%.

Its small size doesn't mean that this camera is weak. The Mini still has the same waterproofing as other GoPros (down to 33 feet). Consider it a license to get wet.