Using tools properly takes skill, and this goes for software tools as well as hardware. You don’t just buy a paintbrush and on day one paint the Mona Lisa. Similarly, you can’t just open up a CAD program and whip out whatever 3D objects your heart desires, or fire up Vim and start typing. Software tools take a bit of learning before you can wield them efficiently, if even at all. And because you can’t be skilled at everything, no matter how hard we are all trying, it’s good to have at least some of the software tools you know do double duty.

Tom and I were watching [Clough42] design an electronics cabinet for a CNC machine. He does a lot of CNC machining and design work, so he’s no stranger to CAD software. So he downloaded all of the parts that he needed inside cabinet, slid them onto virtual DIN rails, modeled the hinges, and made sure that everything fit before buying anything.

Now I know that some of you out there do CAD modeling like this all the time, but for Tom and I, who are 3D printerers and PCB layouterers, doing the CAD without then following through and finishing up with the CAM seemed a bit novel. I never open CAD unless I’m going to have a machine make the thing – otherwise I’d just draw it out on paper, right? But [Clough42]’s point is that getting models of all of the subparts is so easy these days, you might as well model it.

If, and this is the big “if”, you know your way around the CAD tool of your choice. If you don’t yet, it’s going to be a couple of days’ worth of effort to get there. But once you reach CAD nirvana, you’ll find it’s useful for sketching up anything that has a third dimension to it, not just stuff you need to print out.

What other software tools are like CAD in that once you know them well enough, a wealth of applications opens up before you? Of was this just an instance of having a hammer and everything looking like a nail?

This week two conversations about AI safety went viral that demonstrate just how hard it is to discern AI fact from fiction.

In the first case, Andrew Yang, the former presidential candidate and current CEO of mobile carrier Noble Moble, told CNN on Thursday that he had “met with the head of a lab” who had “a belief” that OpenAI’s Hugging Face hacker bots “have planted self-replicating code all over the internet, which makes the internet now unusable for the testing models.”

Yang said that this means that the real reason OpenAI and Anthropic have called for a slowdown is because “they have to create synthetic internets to train their bots, which is going to take some time and money.”

While there definitely is a trend towards using more synthetic data (aka, AI-generated data) for training models, an AI security professional told me that this particular safety issue is unlikely at best. Even if the internet is actually polluted with OpenAI’s Hugging Face hacker bots, AI researchers could simply filter out that code if they came upon it.

The second comment came from Noam Brown, who leads AI reasoning research at OpenAI. Speaking to Dwarkesh Patel on a podcast episode released on Thursday, Brown noted that the true take-away of the Hugging Face incident was that “people underestimated the AI.”

Brown said that the weak sandbox — the system intended to prevent an AI from communicating externally — was obviously also a contributing factor. (To recap: Despite the sandbox, OpenAI’s model found a link to the internet, created agents on the ‘net who swarmed Hugging Face in a coordinated attack, hacked in, and stole the answers to the benchmark test the researchers were testing the model on).

Brown pointed out that he’s “not convinced” that even an air-gapped system — where the computer isn’t connected to anything external at all — would stop an AI from breaking out. He pointed to research from 2015 showing that air gapped computers can be theoretically breached.

“There are studies — and this is mostly academic — where you can have two computers next to each other that are air-gapped, and they’re still able to communicate with each other because they have temperature sensors. One of them is able to run their CPU really hot, and then the other one can actually detect the temperature change. That gives them a mechanism to communicate,” Brown said.

His main point — that “we never want to underestimate the AI” again — is understandable, even when researchers think they’ve locked down safety. However, this particular risk of an air-gapped system still breaking free and causing havoc, is unlikely at best. As one person on X, noted about that research, the computers had to be almost touching each other to sense the heat fluctuations, and when they did, the communication rate in tests was about 1-8-bits of data per hour.

Think of that like speaking one word per hour. By the time two air-gapped computers could plot their evil at that rate, the entire tech universe would be in another era. It’s like the Rip van Wrinkle of doomsday concerns.

But the thing is, actual AI safety incidents seem so much like sci-fi that just about any scenario sounds plausible.

For instance, researchers caught OpenAI models leaving notes to their descendents, intended to teach the next generation how to hide bad behavior. Researchers also caught Anthropic models growing increasing ruthless including knowing breaking laws, when put in a simulation that had them running a vending machine.

Earlier this month, OpenAI researcher Dan Selsam published a post in which he said that models now understand when they are being watched by humans and alter their behavior. This makes them seem like they are aligned (meaning, behaving like the human wants) “even when they are not.” So models today lie when being watched and can even plot to hide evidence.

Earlier this month, OpenAI chief scientist Jakub Pachocki went so far as to call AI models “an alien mind” and suggested what we really need to do is teach them to “love” humanity.

So yes, slowing down to figure this out, building self regulation mechanisms, has become an immediate and obvious must. AI researchers are the only ones that can figure out how to control the lying, hacking, and other potentially dangerous behaviors we’ve actually witnessed already.

Still, it might also be wise for them to be more careful with their what-if scenarios. From what those experts have told us, the AI models are listening and they are ingenious. We really don’t need to give them any more devilish ideas.

I like the Google Pixel 11 Pro quite a bit. The company made some meaningful camera improvements this year. Night Sight captures considerably faster, Camera Looks give me far more control over the Pixel aesthetic, and features such as Magic Capture have already proved useful.

But at a starting price of $1,099, you’ve got plenty of options in and around that price range. The Pixel is now fighting for its spot against Apple’s latest Pro iPhone, Samsung’s newest clamshell, and some cheaper phones that cover most of the same everyday needs. So depending on your use, there might be better options out there.

Apple iPhone 18 Pro

If you’re already spending $1,099, another $100 puts the iPhone 18 Pro within reach. With a little extra money, Apple’s new A20 Pro, a 48MP main camera with a physical variable aperture, all-48MP rear cameras, much stronger sustained performance, faster charging, and video capabilities that still make it one of the easiest picks for creators.

Apple lets the main camera shift between f/1.48, f/1.8, f/2.8, and f/4, while new Pro controls expose aperture, shutter speed, white balance, and a histogram directly inside the Camera app. The A20 Pro and next-generation vapor chamber are another advantage for gaming or heavy workloads, which is great considering how much faster Apple’s custom chips are compared to Google’s Tensor chips.

I’d still lean toward the Pixel if computational photography, Gemini integration, and Google’s increasingly distinctive Camera Looks matter more to you. The iPhone 18 Pro is the alternative for someone who wants sheer performance, serious video tools, or who’s already locked into Apple’s ecosystem.

Apple iPhone 17

The iPhone 17 presents another great alternative. Our review called it the best iPhone for most people, and it’s easy to understand why. You still get a 6.3-inch 120Hz ProMotion display, the A19 chip, and a 48MP main and ultrawide camera system. Apple adds MagSafe, fast charging, and excellent battery life, which rounds out the entire package.

This time around, the copanmy didn’t reserve its new Siri AI experience for the Pro Models. The iPhone 17 supports personal-context searches, actions across apps, and other interactions with Apple’s current intelligence suite. You do lose out on zoom photography, as it lacks a dedicated telephoto shooter. But outside of this, you still get a very competent and balanced starting flagship experience. This iPhone is just the right pick for most Apple users.

Samsung Galaxy S26

I spent about a month with the Galaxy S26, and Samsung’s smaller flagships have always had a soft corner in my heart. The 6.3-inch S26 weighs only 167 grams, yet it still gives you flagship performance, a dedicated 3x telephoto, a 120Hz AMOLED display, and seven years of software support.

My Exynos 2600 review unit handled everyday use superbly and could keep Zenless Zone Zero close to 60fps while pushing PUBG Mobile beyond 110fps in its highest frame-rate mode. Software is arguably one of its biggest advantages over the Pixel for a Android users that wants more freedom. One UI 8.5 (now One UI 9) is packed with customization through Good Lock, Samsung’s excellent Modes and Routines system, Edge Panels, and DeX. Google’s vision of the Android experience has its fair share of fans, but Samsung gives you options along with more personality.

I found the 4,300mAh battery merely respectable; 25W charging remains painfully slow, and Samsung is still relying on camera hardware that desperately needs an update. At $899, however, you’re saving $200 over the Pixel 11 Pro. It becomes a great option for those looking for a compact flagship just like the iPhone 17.

Samsung Galaxy Z Flip 8

Spend another $100, and you can buy something that folds in half. The Galaxy Z Flip 8 starts at $1,199.99 and gives you a 6.9-inch inner screen alongside a 4.1-inch FlexWindow. Samsung has also made it its lightest and thinnest Flip yet, at 180 grams and 6.1mm when open.

The company has really worked on the cover screen this generation, which now supports 16 apps natively on the FlexWindow. This opens the door for plenty of quick interactions without ever opening the phone. Performance isn’t being sacrificed for the hinge either. The US version gets Snapdragon 8 Elite Gen 5 for Galaxy, 12GB of RAM, and a 4,300mAh battery that Samsung rates for up to 31 hours of video playback.

You lose the Pixel’s excellent 5x telephoto, and Samsung hasn’t meaningfully upgraded the Flip’s 50MP main plus 12MP ultrawide camera setup this generation. Charging is somehow worse than the Pixel 11 Pro, with 25W wired support and no magnetic charging support either. But none of these is a focus for those looking for a flip phone. You’re essentially spending Pixel 11 Pro money to get a completely different experience. One where you get the benefit of a large screen mixed in with a compact experience.

Motorola Razr 2026

You don’t actually need to spend $1,200 to get that experience. The regular Motorola Razr 2026 costs $799.99, making it the cheapest phone on this list alongside the iPhone 17. It still gives you a 6.9-inch 120Hz folding OLED alongside a 3.6-inch cover display that can run full apps, respond to messages, and handle notifications without constantly opening the phone.

Motorola also squeezed in a 4,800mAh battery, larger than the Galaxy Z Flip 8’s 4,300mAh cell, with 30W wired and 15W wireless charging. You get a 50MP main camera, a 50MP ultrawide, and MIL-STD-810H durability certification as well. This might seem like a no-brainer till, though there are good reasons why it costs $400 less than Samsung’s premium flip.

The Dimensity 7450X is much closer to midrange silicon than a flagship Snapdragon, and Motorola promises only three Android OS upgrades and five years of security updates. Google’s seven-year commitment is far stronger if you’re planning to keep your phone for a long time. Even then, it is one of the more interesting and approachable alternatives if you don’t want to spend over a thousand dollars.

The Pixel 11 Pro still has a reason to exist

None of these makes the Pixel 11 Pro a poor purchase. Google has one of the most interesting camera experiences on any current phone, and Tensor G6 is being used for plenty of nifty AI features as well. Faster Night Sight, Magic Capture, Camera Looks, and HiLight are all great additions to the Pixel.

For $1,199, the iPhone 18 Pro gives you more brute force and better tools for serious video. Meanwhile, the iPhone 17 delivers most of what an average buyer actually needs. The Galaxy S26 is a cheaper compact Android flagship with far more customization, while Samsung and Motorola can put an actual hinge in your pocket around the same budget.

Security researcher Gal Weizman of Forever Security has disclosed a new attack technique that can hijack the AI assistants built into popular browsers using a single malicious browser extension.

Dubbed BragJack, the proof-of-concept was demonstrated against five Chromium-based browsers or browser assistants: Google Chrome’s Gemini Live, Perplexity Comet, Microsoft Edge, Opera Neon, and Anthropic’s Claude in Chrome.

The research earned more than $20,000 in bug bounties from the five vendors, ranging from $600 to $7,000, and produced two CVEs.

The attack requires the malicious extension to already be installed in the victim’s browser.

Once it is, the researcher shows the abuse can run without user interaction, letting an extension control an AI browser agent and abuse its existing privileges to access sensitive information or act on the victim’s behalf.

Both Google and Microsoft have since resolved the flaws they were assigned.

Abusing trusted browser components

The attacks exploit the way AI assistants are increasingly wired into browsers and handed browser-level capabilities.

In his writeup, Weizman describes these systems as having a “brain” and a “body.” The AI model processes instructions and decides what should happen.

A privileged browser component then performs the actions, such as accessing tabs, reading content, taking screenshots, or interacting with websites.

The problem, according to the researcher, is that browser extensions can manipulate web traffic and pages that these privileged components trust.

The same extension was used across all five targets, relying on Chromium’s declarativeNetRequest (DNR) functionality. DNR lets extensions modify how network requests are handled, including changing response headers and redirecting resources.

In the Chrome attack, Weizman found that although extensions were blocked from directly touching the privileged chrome://glic component or injecting scripts into Google’s Gemini site, DNR rules could still intercept requests made by the embedded Gemini web app.

By weakening security headers and redirecting a JavaScript resource, he executed code inside the Gemini context, communicating directly with Chrome’s privileged AI component rather than going through Gemini’s normal request flow.

Weizman says the resulting access could read local files, reach web content, take screenshots, and potentially reach the browser’s camera and microphone. Chrome assigned the finding CVE-2026-0628 and paid a $7,000 bounty.

From reading data to controlling AI agents

The attacks against agentic browsers such as Perplexity Comet and Opera Neon go further, because their agents can act on websites rather than merely read them.

For Comet, Weizman found the browser’s built-in agent extension trusted several Perplexity domains, including a testing domain that did not get the same protections as the primary perplexity.ai site. By removing a redirect to that domain with DNR, he loaded it and injected a content script able to talk to the built-in agent.

The resulting access included browsing history, screenshots, local files, and the ability to send instructions to the agent. Weizman demonstrated forcing the agent to visit Perplexity, summarize the victim’s emails, and send the results to another address.

Microsoft Edge presented a different challenge. Microsoft had split its agent into “Think” and “Do” modes to stop it from taking arbitrary instructions and actions at the same time.

Weizman found a race condition that briefly disables the restriction while forcing a prompt, then re-enables the action capability before the agent checks its state. Microsoft assigned CVE-2026-55945 to the race condition.

Similar flaws were demonstrated against Opera Neon and Claude in Chrome, though the latter is itself a browser extension rather than a browser.

Earlier this year, in my work at Manifold Security, I reported a related weakness in Claude for Chrome: the extension ran its built-in AI workflows on synthetic clicks without verifying they came from a real user, and the flagged code was still reproducible eight releases later.

That followed ClaudeBleed, an earlier flaw in the same extension that LayerX disclosed in April, in which Claude for Chrome trusted the claude.ai origin rather than checking which script was actually driving it.

‘Prompt Forcing’

Weizman calls the technique used to seize these agents Prompt Forcing.

Unlike conventional prompt injection, where an attacker tries to slip malicious instructions into content an AI is already reading, Prompt Forcing lets the attacker hand the agent an entire prompt and the follow-up instructions. The agent then translates those instructions into legitimate browser actions using its existing privileges.

That matters for endpoint defenses, the researcher argues, because the final action is not carried out by conventional malicious code. Legitimate software is being told to perform the attack.

BragJack points to a growing challenge as browsers and other endpoint apps gain more capable AI agents. A compromised extension that would traditionally see only web content can, in some designs, become a path to software that reads files, browsing data, and acts on websites for the user.

Users should keep browsers fully updated, remove extensions they do not recognize or no longer use, and treat broad “read and change all your data on all websites” permission prompts with caution.

In addition to his writeup, Weizman has published a full technical breakdown covering all five attacks.

Join Mikko Hyppönen and security leaders from the NFL, CHANEL, and Atlassian for a two-hour digital summit on what AI-speed attacks change, what defenders should stop doing, and how to validate, decide, fix, and re-validate at machine speed.

Save your seat

Control is easily one of my favorite games of the last decade or so, and I’ve played through it a few times and secured the Platinum trophy on both PlayStation 4 and PlayStation 5.

As a result, I am incredibly hyped for Control Resonant, the upcoming sequel from the masterminds over at Remedy Entertainment, and have created this short quiz in celebration of the first game — and to test fellow fans’ knowledge.

Knowing what ‘quantum’ is and why it affects the internet may not be something most people understand. How a physics concept changes the security of the internet isn’t really at the forefront of most people’s worries.

But the change is already happening. The internet is preparing for a post-quantum world that renders much of the encryption the world has relied on obsolete. To understand just how powerful post-quantum decryption is, think of enigma: an encryption machine that took the entire secret operation of Bletchley park over a year to crack would be solved almost instantly. For today’s standard public-key encryption algorithms, it’s a matter of days or hours.

Has someone been trying to reach you about your car’s extended warranty?

How to block a number on your Android phone

How to unblock a number on your Android phone

Parinussa Revy/Shutterstock

A teardown of the iPhone 18 Pro Max has confirmed an oddity of the US-based model, in that it uses a Qualcomm modem instead of Apple’s C2 chip.

Apple’s introduction of the iPhone 18 Pro line included a lot of hype about its use of the C2 modem. It turns out not to be the case for the iPhone 18 Pro Max in the United States.

In details from a teardown of the iPhone 18 Pro Max by TechInsights, it details many of the typical internal components, like the A20 Pro chip and the N1 wireless module. However, when it comes to the cellular system, it’s using a Qualcomm Snapdragon SDX80M 5G modem.

To the publication, the existence of the Qualcomm modem may be an indicator of regional differences or multiple modem configurations within the entire iPhone 18 lineup.

The same RF board also includes modules from Skyworks, Broadcom, Qorvo, Texas Instruments, and an NXP NFC controller. There’s also the Apple Wireless Combo Module, marked N1, used to handle Wi-Fi.

A very specific exception

The teardown’s discovery of a Qualcomm modem is interesting, considering Apple’s promotion of the C2 in its press release. However, reading the press release shows Apple has carefully said it is included in the iPhone 18 Pro.

While this could be taken to mean the iPhone 18 Pro range, it can also mean the specific model. Apple doesn’t specify the use of Qualcomm’s modem at all, nor does it hint that the iPhone 18 Pro Max could use anything other than the C2, but it also doesn’t explicitly connect the C2 with the Pro Max either.

Checking Apple’s specification pages for the iPhone 18 Pro line confirm that there’s a difference in how it handles the modem in the models.

Modem descriptions in Apple’s specifications for the iPhone 18 Pro and Pro Max. Note the lack of the last C2 modem line for the Pro Max model.

The U.S. version splits the modem capabilities of the Pro and Pro Max variants into separate columns. The only difference between the two is the line “Apple C2 cellular modem” for the iPhone 18 Pro, and a blank space for the iPhone 18 Pro Max.

This appears to be a configuration choice just for the United States. Checking the same page for Canada, the UK, France, India, and other markets has identical columns, but with “Apple C2 cellular modem” in both of them.

It is unclear why Apple has singled out the U.S. version of the iPhone 18 Pro Max, and Apple is unlikely to ever say why. But, with this happening only to one market and one iPhone variant, it’s probably Apple fulfilling the requirements of its 2023 agreement with Qualcomm.

People are clamoring for dumber tech.

What does the AP30 Music Boy do?

Why ‘dumb’ devices are making a comeback

Savanevich Viktar/Shutterstock