AI safety for both adults and children is under greater scrutiny than ever before, with a growing number of questions and concerns appearing around how beneficial or otherwise these conversational bots actually are for our mental well-being—even without considering their energy use, ethics, or economic implications. At the same time, an increasing number of apps and services are becoming age-gated in a growing number of countries around the world, as governments realize that letting kids loose on AI apps, social media platforms, and the web at large might not be the best idea for their ongoing development and prosperity.

The big AI apps, such as ChatGPT and Gemini, are right in the center of these trends, and the companies behind them are taking steps that they promise will keep children safe. If you want to know what youngsters in your family may be getting exposed to, and what you can do about it, here’s the rundown.

ChatGPT

OpenAI’s ChatGPT has one of the more comprehensive setups in terms of parental controls. The minimum age for users is 13, and if you’re under 18, you need to have parental consent to open an account—age is predicted via several signals, including what times of day you use ChatGPT and what you talk about.

Accounts for under-18s automatically become ChatGPT for Teens accounts. There’s more of a focus on learning and study (with a dedicated study mode), as well as “age-appropriate safeguards to help reduce exposure to sensitive or potentially harmful content.” There are built-in protections for “harmful topics,” including sexual, romantic, or violent roleplay, graphic violence, and self-harm. These work on top of the safeguards for all users that limit chats about physical harm and illegal activity.

These ChatGPT for Teens accounts can also be linked to a parental account—but while parental consent is required for under-18s to use the service, ChatGPT clearly says that teens do have the option to break the link if they want. If you do persuade your kid to keep your accounts connected, you can set time limits on when the AI bot can be used, and tweak some of the sensitivity settings. However, you can’t view any conversations.

To set up parental controls from ChatGPT on the web, click your profile icon (bottom left), then Settings > Parental controls > Add family member and follow the instructions. It’s also worth bearing in mind that ChatGPT app can also be used for free, anonymously, in any web browser, though OpenAI says users accessing the AI this way are subject to the default under-18s limits (and can’t save chats or access an archive).

Gemini

Gemini is tied into everything else Google does, and follows the broader rules for Google accounts: No under-13s, unless you’re part of a family group and using an account managed by an adult. You can manage your family through the Family Link portal on the web, though as far as parental controls specific to Gemini go, you can only enable or disable access to the AI (settings like screen time apply across all enabled apps).

For all under-18s (partly identified through age estimation), whether or not they’re part of a managed family group, Google applies some additional protections. They include preventing the AI from taking on a persona, guarding against emotional dependence on the AI, and discouraging bullying and other forms of harassment. There are also safeguards to limit chats about age-restricted substances like alcohol or vapes, and mental health issues like eating disorders.

These protections for minors are applied in addition to the usual guardrails in Gemini for everyone. The guardrails should lead you away from conversations about self-harm, suicide, violence, and illegal activities, and offer quick links to support in times of crisis. However, Google itself admits that these filters aren’t perfect.

As with ChatGPT, there’s a loophole of sorts in that anyone can start chatting with Gemini, for free inside a web browser, with no checks at all. These conversations can’t be saved and come up against usage limits fairly quickly, but they’re only protected by the general restrictions for all accounts, not the ones specific to minors.

Meta AI

Meta AI will show up in several apps that young people and children are likely to use, including WhatsApp and Instagram, as well as having its own separate app on the web and on phones. The standard restrictions for Meta’s social media accounts apply to its AI too: A minimum age of 13, and restrictions for all under-18s. Like many other platforms, it uses some AI magic to flag up accounts being used by minors.

If you’re under 18 and on a Meta Teen account, you’ll come up against blocks if you try to get Meta AI to engage in sexual or romantic roleplay, or discuss alcohol or substance misuse. Chats on topics such as eating disorders and self-harm will prompt the AI to offer guidance towards relevant support services.

Meta offers a comprehensive parental control setup for its apps, including Meta AI, but connecting a parent account isn’t compulsory for teens. This is handled through the Meta Family Center portal, and parents can get reports on the general topics that kids are chatting about with AI—though they don’t get the exact transcripts or prompts. Any self-harm or suicide discussion can trigger an alert to the connected parent account.

All users, under 18 or over 18, get support and crisis intervention responses when chatting about suicide and self-harm, and there are strict limits on discussing illegal and harmful acts, including drug use and cyberbullying, as well as sexual and adult content.

Siri AI

Apple users have to be 13 or over to have an account, which is required to access Siri AI—anyone younger needs to have an account created for them by an adult, and these accounts can only get at the basic, older Siri. Date of birth must be declared when opening an Apple account, with tighter checks rolling out in many parts of the world.

Anyone over the age of 13 can have an Apple account and use the Siri AI app, with or without a connected parent account, as it stands in most U.S. states. This means the well-being safeguards are essentially the same for teens as they are for adults: There are safeguards around self-harm, suicide, sexual and violent content, and illegal commands. It’s one of the more conservative AIs when it comes to what it will talk about.

The new and improved Siri AI can be managed to some extent by the new parental controls available in iOS 27 and Apple’s other software updates for 2026—Apple has clearly seen the way the wind is blowing—but this is optional for teens. It depends on whether or not you can corral your kids into a Family Sharing group.

If your teens are in your family group of users, then the standard Screen Time rules apply for controlling app usage limits and device access (in iOS Settings, tap your name, then Family to get at them). There are also options for disabling certain Siri AI features, such as the writing tools or the image creation capabilities.

Copilot

Like Gemini/Google and Apple/Siri, Copilot is tied into the wider Microsoft ecosystem, and access for under-18s can be managed through it. Users must be 13 or older to hold a Microsoft or a Copilot account—and again, there’s some underlying tech in place to detect anyone who might be deliberately lying about their age.

If you’ve got a teen using Copilot, you can manage their access through the standard Microsoft Family Safety portal. You’re able to enable or disable Copilot access, set times during which Copilot can and can’t be used, and disable certain features like image generation or voice mode. Parents can’t read the Copilot chats of their kids, though.

Parental supervision is optional—a teen aged 13–17 can use Copilot without linking their account to an adult one. Microsoft says users identified as under 18 won’t see personalized results or ads, and will get age-appropriate responses regarding risky or harmful content, though there’s not much detail on what that entails.

More generally, content around self-harm, violence, adult content, harassment, and harmful or illegal activity is banned on the AI platform. Any prompts about suicide or drugs will bring up support options rather than answers, whether you’re an adult or a young person using the service.

Grok

Out of the bigger AI chatbots out there, Grok is very much the worst at protecting minors. It can be accessed through X or as its own standalone app, and while its terms of use indicate there’s a minimum age requirement of 13, a recent report by Common Sense Media says the platform’s age checks aren’t effective.

Ultimately that means that minors have the same access to Grok as adults, if they want it—an AI platform that combines “weak safety guardrails” and “a range of extremely harmful content” that includes “explicit sexual content, mental health risks, violent content, and inappropriate responses to… offensive content” (that’s the Common Sense Media Report again).

Yes, this is the AI chatbot that has previously had a problem with deepfake porn imagery, and which has traditionally taken a rather casual attitude towards safety concerns, led by SpaceXAI CEO Elon Musk. While illegal and harmful content is supposedly banned from Grok, its own FAQs admit that certain chats can result in “dialogue that may involve coarse language, crude humor, sexual situations, or violence.”

There is actually a Grok Kids Mode that limits the kind of content that can be shown and the conversation topics that can be added, but it requires users to self-declare themselves as minors and is easy to bypass. The advice from Common Sense Media seems to be the way to go with Grok: Do not let your teens access it.

Claude

Claude’s approach to child safety is straightforward: You can only register for an account if you’re over 18. It is possible to lie about your age when signing up, but like the other platforms, Anthropic looks at various “signals” in behavior to spot minors—if your account gets flagged, you have to verify your age using an option like face age estimation or Yoti.

There’s no free web access to the chatbot, like there is with ChatGPT and Gemini, so if you’re under 18, you’re locked out from the service unless you know the behavioral tricks to get past Claude’s detection systems.

In terms of more general well-being protections on Claude, any chat around self-harm or suicide will guide users towards helplines and other resources for support. Conversations about illegal activities, disallowed speech, and explicit sexual content are banned from the platform. In recent months, Anthropic says it’s also reduced AI sycophancy.

That’s something of a whistle-stop tour of the big seven AI chatbots, but if you’re a parent, it should give you some idea of what’s out there and how easy it is to access. Besides the parental controls built into these AIs (or not), you also have the settings available at the platform level—Android, iOS, Windows, macOS—to restrict the use of any potentially harmful apps.

LEGO’s BrickLink marketplace is back online after a week-long outage caused by what the company described as a “cyber incident,” during which an unauthorized third party accessed some user data. Details about exactly what data was accessed remain unclear. The site went offline on November 5 and remained unavailable until November 12, preventing users from buying or selling LEGO sets and parts during the busy holiday shopping season. According to Brick Fanatics, LEGO initially described the outage as “necessary work” before confirming that the platform had suffered a security incident.

In an email to users, LEGO acknowledged that “some of the information associated with your account was accessed and disclosed by an unauthorized party,” including potentially usernames and IP addresses. However, the company insisted that “your password, postal address, and financial information were not disclosed.” Those assurances have since come under scrutiny following reports of convincing phishing emails containing detailed personal and purchase information.

The outage also dealt a significant blow to independent sellers who rely on BrickLink for income. Simon Webster, who operates stores on both BrickLink and rival marketplace Brick Owl, reported that his overall sales revenue fell 52% over a four-day period, even as sales through Brick Owl rose 79%. “Their comms are amateur at the best of times, but surely they should be able to send an email to let sellers and buyers alike know what is happening and what they’re doing to rectify it,” Webster said. “But no, we’re all left in the dark.”

The incident has renewed concerns about BrickLink’s aging infrastructure. Founded in 2000 and acquired by LEGO in 2019, the marketplace reportedly runs on a complex collection of interconnected systems that makes modernization difficult. The Brick Fanatics report questions whether the 25-year-old platform remains fit for purpose and whether prolonged outages and security concerns could push sellers toward competing marketplaces like Brick Owl.

Read more of this story at Slashdot.

On Friday afternoon, days after expanding use of an AI tool, the Federal Aviation Administration issued a nationwide ground stop, halting departures from all US airports. Secretary of Transportation Sean Duffy announced the ground stop on X, citing a “communication issue impacting all Air Route Traffic Control Centers.” Duffy promised updates as more information is known and said FAA staff were troubleshooting the problem.

Department of Homeland Security Secretary Markwayne Mullin said that the ground stop was not prompted by a security threat. “This is a telecommunications outage out of the FAA. Out of an abundance of caution, we have grounded all of the flights,” Mullin said. “There’s no indication as of yet that this is a nefarious attack against us.”

Mullin said the FAA would handle incoming international flights on a case-by-case basis.

Read full article

Comments

The box was already for decoration, and now maybe more so.

We may earn a commission from links on this page.

The latest Apple Watches have a lot of new features that older watches don’t, so I figured it was time for a new list of hacks. While our general Apple Watch and fitness tracking hacks are still golden, Series 12 devices offer new features like native step counting, a Readiness app, an upgraded heart rate sensor, an AI-based version of Siri, and “audio intelligence” that can replay and summarize real-life conversations, within limits. Here’s how to get the most out of them.

Put the steps complication on your watch face to turn it into a simple pedometer

The Series 12 can finally count steps natively, a task that for some reason is only arriving now, and is restricted to just the newest Apple Watch models. Enjoy it! To add a step counter to your watch face, long-press on the clock screen to edit it, tap on a complication (that’s what Apple calls those little icons and numbers that appear on some watch faces), choose Activity, and then tap Steps. 

Get better Siri Recaps by turning them off by default

You don’t need this creepy feature listening in to everything, but it can be nice to have it as a backup on occasion. A button in your Apple Watch’s Control Center can turn the feature on and off. Keep it off except when you have a structured conversation that you’ll want to refer to later. This feature is currently in beta (which you can try now if you want to), but is expected to be available to everyone later this year.

It’s worth turning off the default setting not just because of the creepiness factor, but also because the recaps work best when there is a single subject of conversation, with a clear start and end, and just one context. If you leave recaps running all the time, your doctor’s appointment will get mixed in with your grocery shopping, YouTube watching, and everyday chatter. 

To finish off a recap and make it useful, take a look at the recap shortly after the conversation, tap the three-dots menu, and hit Edit. You’ll be able to adjust the notes and add your own important details. Siri won’t identify speakers and seems to shy away from including certain details, like people’s names. You can add in what it missed and also create your own bullet points so that everything gets noted down while it’s still fresh in your mind. 

Learn these gestures to access features more easily

This one isn’t specific to the Series 12, but if you haven’t gotten on board with the pinch and wrist flick gestures yet, now is a great time. 

First, the wrist flick: If a notification is on your screen, or you’re otherwise looking at something you’d rather be done with, just flick your wrist away from you and then back toward you, as if there were a crumb on the watch face and you wanted to knock it off. This will bring you back to the watch face and close whatever you were looking at. 

The double pinch is another useful one, although Apple calls it double tap. This is confusing because you aren’t double tapping the watch, you’re tapping your thumb and forefinger together while wearing the watch. You don’t need to touch the watch face at all. So, when you get a notification or a call, look at it on your watch, and pinch your thumb and forefinger (of the watch hand) together twice. The watch detects the pinch from subtle movements in your wrist, and it will open the notification. You can then use the wrist flick to close the notification. Performing a double pinch when no notification is on screen will bring up your smart stack. You can then double-pinch to move to the next card in the smart stack, or single-pinch to open that card and see its details. 

If you want to learn even more gestures, go into the Watch app, and select Accessibility and then Hand gestures. You can remap a whole-fist “clench” to do the same thing as pressing the digital crown, for example, or to open the control center or Siri. 

Slap your watch to make Siri shut up

A little rude, perhaps, but helpful: With the new AI-based Siri, sometimes answers to questions can be long-winded. If Siri launches into a novel-length response and you’re not interested in hearing all that, just put your palm over the screen. You’ll instantly be transported back to your watch face, and the Siri response will stop. 

Turn on Silent responses so you can read what Siri says instead of listening to it

If you still want to see those longer responses without having Siri read the whole thing to you, you can set this up on your phone, and the settings will carry over to your watch. On your iPhone, go to Settings and tap Siri, then Responses. Select Silent responses instead of Spoken responses. You can then read through her answers, or skim them, as you prefer.

Mozilla is shutting down Solo, its AI-powered website builder and hosting service, on November 30. A notice on the Soloist.ai homepage reads: “We’ve made the difficult decision to close Solo. We have appreciated all of your support building and growing this product, and the feedback we’ve received from all of you. Solo will cease operations on Nov 30, 2026.” The Register reports: Mozilla launched Solo in December 2023 as a “Mozilla Innovation Project,” aimed at helping sole traders to build themselves a webpage for their services. Some 54,000 people used its LLM to build pages, and relied on Mozilla to host them — complete with custom domain names. Users now have just over two weeks left to find a new home for their pages, although Mozilla says it first notified them on October 1. According to the support FAQ, users can download a ZIP containing an HTML export, contact form and newsletter submissions, and a CSV linking to their images. The images themselves must be downloaded separately. Domains purchased through Solo Domains must also be moved to Name.com, with Mozilla handling the transfer after users create an account and supply its code.

Read more of this story at Slashdot.

Elephants live in some of the most complex and resilient societies in the animal kingdom. Individuals die or move away, calves grow up, and newcomers join, but the local communities as a whole usually hold together. Long-term studies of African savanna elephants have shown that, even after poaching or other spikes in deaths, survivors tend to rebuild their lost connections over time. But a new study suggests that resilience may not keep the community intact when steady losses accumulate over time.

Anastasia Madsen, a behavioral ecologist at the University of California, San Diego, and her colleagues analyzed 12 years of observations of wild Asian elephants in Sri Lanka. They found that, once losses began to outpace newcomers, the elephants’ society didn’t just get smaller. It fractured.

Accidental savanna

Udawalawe National Park in southern Sri Lanka was created in the 1970s around a reservoir on one of the country’s four major rivers. “Before that, it was basically a tropical deciduous forest,” said Shermin de Silva, an ecologist at UC San Diego and senior author of the study (de Silva started the Udawalawe Elephant Research Project in 2006).

Read full article

Comments

You probably already know that your email isn’t actually all that private, especially if you have a free account with a major service provider like Gmail (Google) or Outlook (Microsoft). But if you are sending sensitive information via email, you can minimize what can be snooped on and by whom using end-to-end encryption, which generally ensures that only the recipient can read your message. Here’s what you need to know about email encryption and how to set up an extra layer of privacy.

How email encryption works

As this in-depth explainer from Adaptive Security describes, most emails are encrypted in transit via Transport Layer Security (TLS), which prevents anyone from reading your message as it travels from server to server. However, encryption is temporarily stripped at each server: content is received and processed in plaintext and then re-encrypted when it leaves, meaning it could be read, logged, and distributed by the server provider.

Emails are also typically encrypted at rest—for example, while sitting on Google’s servers and in a Gmail recipient’s inbox. But this only prevents someone with physical access to the server hardware from reading messages, as Google still holds the encryption key and can access stored data. And, as PCMag points out, Google is scanning your messages to a degree, and it (along with other email providers) can and will hand emails over to law enforcement and government agencies in some cases.

By contrast, with end-to-end encryption (E2EE), emails are readable only on the sender’s and recipient’s devices, and can be decrypted only with the recipient’s private key. They’re never translated into plaintext as they pass between and through servers to their destination, meaning that no one—including server providers, hackers, or law enforcement—can view the content. It’s worth noting, though, that E2EE is not impenetrable nor a total guarantee of privacy. For example, metadata like subject lines and email addresses may remain in plaintext, and encrypted messages are still susceptible to phishing. Plus, forwarding an encrypted message breaks the encryption.

Still, end-to-end encryption is especially useful when sending sensitive information that may be valuable to bad actors, though if security and privacy are a high priority, you may want to consider it for all of your emails. How you deploy encryption depends on whether you are adding it to your current email or switching to a dedicated provider.

Add third-party encryption to your existing email

If you have an account with a major email provider like Gmail, Outlook, or Apple Mail, you can add a third-party tool that layers an encryption protocol like OpenPGP on top of your existing account. PCMag recommends PreVeil, SecureMyEmail, and Virtru (which is limited to Gmail users in Chrome) for those who want to keep their current emails. Mailvelope is a browser add-on (and therefore a desktop-only tool) available in Chrome, Firefox, and Edge, though it can also be installed on other browsers, like Brave, that are built on the same architecture. FlowCrypt integrates with Gmail and Google Workspace on desktop and mobile.

Note that enterprise users of Google Workspace and Microsoft 365 have access to true end-to-end encryption without third-party add-ons. Gmail’s Confidential mode and Outlook’s Encrypt feature, available to all users, are not E2EE equivalents, as Google and Microsoft, respectively, still manage the keys.

Switch to a dedicated encrypted email provider

For those who are willing to start fresh with a new email address (or who have a custom domain), providers like Proton Mail or Tuta offer native end-to-end encryption. The caveat: by default, this is only between users on the same service, so from one Proton user to another. If you want to send an end-to-end encrypted email from Proton to a Gmail user, for example, you have to set up password protection (and share the password via another secure channel) or use a PGP-compatible tool like those listed above. You can also Proton’s Gmail integration does allow for end-to-end encryption, but only if both Gmail parties are using the Proton integration.

The bottom line here is that email wasn’t designed to be a highly secure form of communication. End-to-end encryption does add a layer of privacy and security if you jump through hoops to set it up, but it’s still susceptible to human error and social engineering.