News

• Chinese research vessel lingered above critical Pacific undersea communications infrastructure
• Taiwan's Coast Guard warned the vessel to turn away from the Pacific Light Cable Network
• Windward linked slow vessel movements with recognized intelligence-gathering behavior

A Chinese-flagged research vessel was recently observed loitering directly above an undersea cable system connecting Taiwan to the United States.

Taiwan's Coast Guard filmed the encounter and warned the 200-foot ship to alter its course away from the cable route.

The cable is the Pacific Light Cable Network, an 8,000-mile fiber-optic system linking Taiwan, the Philippines and the US.

Maritime analysis platform Windward tracked the vessel moving at slow speed directly over the cable route in the Philippine Sea.

The company described such slow-speed movement over critical undersea infrastructure as one of the clearest signals used in maritime intelligence gathering.

Windward stated plainly that cables do not move, adding that "vessels that hover over them get noticed" by trained analysts worldwide.

Though the vessel's name and identifying number were withheld, Windward said it had operated repeatedly near the disputed Paracel and Spratly Islands.

Records also placed the same ship as far as the Bay of Bengal and waters off Sri Lanka in earlier years.

Analysts also noted increased Chinese research vessel activity east of Taiwan since June alongside broader maritime operations in surrounding waters.

Some analysts have expressed concerns that Beijing's military-civil fusion policy could enable information gathered by civilian research vessels to support military activities.

However, no publicly available evidence has verified that the vessel gathered intelligence or transferred information to Chinese military authorities during the reported voyage.

Chinese authorities have consistently denied that their vessels intentionally damage undersea cables, calling such accusations premature and politically motivated overall.

They accuse Taiwan and several Western governments of concluding before formal investigations into these incidents are completed.

Chinese-flagged vessels have featured in previous investigations into damaged undersea cables around Taiwan.

NATO officials and Taiwan's own Coast Guard have also linked Chinese-flagged vessels to earlier cable disruptions, including suspected sabotage in the Baltic Sea.

On its part, China claims that all its maritime activities are completely legal and respects international maritime law.

Submarine cables worldwide now carry roughly 99% of all intercontinental internet traffic, making their protection a growing global security concern.

The Pacific Light Cable Network carries added significance because Taiwan alone produces more than 90% of the world's most advanced semiconductor chips today.

This network was originally planned to connect Hong Kong before US regulators blocked that route in 2020.

American officials cited concerns at the time that the connection could directly expose sensitive US data to Chinese intelligence services.

The project was later reconfigured to link only the United States, Taiwan and the Philippines, bypassing Hong Kong entirely afterward.

Analysts believe the recent deployments may gradually normalize China's presence in waters that Taiwan has long regarded as strategically important.

Whether the vessel's activity ultimately amounted to scientific research or intelligence collection remains unresolved.

Via Newsweek

To be online is to expose yourself to numerous threats, and those aren’t limited to hackers and viruses; data brokers, free services where you pay with your information, and — depending on your situation — even your own government can be a concern too. And when your whole online life is accessible from a phone in your pocket, there’s also the real threat of a thief getting into it.

With all that in mind, I’ve always been somewhat security conscious, but I’ve also historically been a bit inconsistent about my security practices, and tended to prioritize convenience over safety. A while back, though, I started taking things more seriously.

I started listening to podcasts like Firewalls Don’t Stop Dragons, digging into privacy and security subreddits, and trying out numerous new products and services to beef up my digital defenses.

But let’s look at where I started first. Before embarking on this journey, I already used both Norton as my primary desktop antivirus and Malwarebytes as an on-demand scanner. I also used a password manager (1Password) across both phone and desktop, and ensured all my passwords were strong and unique.

On my iPhone, I had Stolen Device Protection enabled, which makes it harder for thieves to access sensitive information or change my Apple account password.

And I had two-factor authentication set up for everything that would allow it — opting for one of the best authenticator apps over SMS, since the latter is vulnerable to SIM swap attacks, where hackers convince your mobile network to port your number to their device.

I still do all those things now, and would recommend most other people do too — though when it comes to the best antivirus and best password manager, there are several other options to consider as well. But I’ve also added other services and even devices to my security toolkit.

There’s a saying that “if something is free, you are the product," and that’s largely true — and can have massive privacy implications.

Did you know, for example, that Gmail and Outlook can use AI or other automated systems to scan the content of your emails? And while the companies claim they’re not read by people, in theory it’s possible for them to be, either by the company itself, or by law enforcement if they’re ever served a warrant.

But you don’t have to be committing crimes to feel uneasy about the idea of an AI essentially reading all your supposedly private communications.

And this issue isn’t limited to email — the same privacy concerns apply to most free cloud storage services, whether we’re talking photo storage things like Google Photos, or general purpose ones like Dropbox.

Some services are worse for this than others, but if they’re not end-to-end encrypted, then your files, photos, and missives can, in theory, be viewed by people other than you and any recipient. That means, even if you’re not using a free service, you might still be making privacy compromises if the company has other priorities, though free options tend to be the worst for this sort of thing.

So with that in mind, I signed up with Proton Unlimited, giving me end-to-end encrypted email through Proton Mail, and cloud storage through Proton Drive.

As a bonus, Proton Unlimited also comes with a capable VPN and password manager in Proton VPN and Proton Pass. I use the former (more on this below), but not the latter, as personally, I prefer 1Password, and also don’t love the idea of having my whole security infrastructure handled by one brand.

This hasn’t been a complete switch, as I do in fact still use my less secure emails and cloud storage for some things, but sensitive stuff like bank logins have all been switched to Proton, while things like encryption keys are backed up in Proton Drive.

Part of why I haven’t fully switched is just the hassle of moving everything across, but encryption does come with some downsides. For example, it makes the search function more limited, as even the app can’t read the content of documents or other files.

And there are upsides to keeping alternative email addresses, as I can use them for things that I don’t trust not to spam me, or for accounts I don't trust not to get hacked, while my more important logins and communications are all kept secure at another address.

Though on that front, Proton does have the handy ability to let you create unlimited hide-my-email aliases, so I don’t strictly need a whole separate account for this.

In any case, while I recommend Proton as the best secure email provider, there are other options too, like Tuta and Mailbox.org, and many of these do have free plans too, though those are usually quite limited.

My privacy journey also led me to limit what traces of myself can be found online, and the big thing there, of course, is a VPN.

Technically, I wasn’t new to VPNs, but in the past I mostly only used them to get around regional restrictions. Now, I have a VPN active most of the time. This hides my location from websites, since they don’t see my real IP. It also hides my browsing activity from my ISP, and it keeps my passwords and other private information more secure when using public Wi-Fi, so it has a lot of upsides.

The main downside, as I see it, is that some sites block VPNs, so you potentially need to disable it at times if you want to access those sites. However, they don’t always block every VPN server, so simply switching servers can solve that issue.

If you’re on a device that supports split tunnelling, then you could also enable this — allowing you, for example, to have a secondary browser that doesn’t connect via VPN for sites where it’s an issue, while everything else you do does. As noted above, I use Proton VPN, but there are other great options too in our best VPNs guide.

But VPNs are just one part of reducing your digital footprint. You can also choose to keep your social media anonymous or private — something that I was already largely doing — and you can set about removing the data that companies already have on you.

I took two approaches to this. The first was subscribing to Incogni. This is a data removal service that will send removal requests to hundreds of data brokers. These brokers likely have all sorts of data about you, linking your name with your phone number, email address, and more, which makes you an easier target for spam, scams, and identity theft.

Incogni isn’t free, but it has been independently assessed by Deloitte, so you can be confident that it’s doing what it claims to do. And while you could manually make data removal requests to all these companies, it would likely take a very long time.

Of course, while I rate Incogni, it’s not the only option — check out our best data removal services guide for other solutions.

The final step I’ve taken for reducing my digital footprint is removing my face from facial recognition search engines like PimEyes and FaceCheck.ID. This I did do manually, though, by following their respective opt-out processes, which basically involve uploading a clear photo of your own face and asking for that face to be removed. You can find the exact process on the PimEyes opt-out page and the FaceCheck opt-out page.

I mentioned above that I previously favored an app — specifically Authy in my case — over SMS for two-factor authentication, but as part of my privacy upgrade, I went a step further and invested in some YubiKey security keys.

These are hardware keys that are basically tiny USB sticks that you put into your phone or computer to perform two-factor authentication, and you can also get NFC versions with no plugging required.

Unlike apps or SMS, which are vulnerable to phishing attempts or hacks, YubiKeys (and similar devices from other brands) need to be in your physical possession to use. So they’re safe from online threats. Of course, someone could physically steal your YubiKey, but they’d still need your account passwords as well.

So if you want the strongest possible two-factor authentication, I’d recommend investing in one of the best security keys.

However, this is actually the security upgrade that I’ve found myself using least, as I simply find it less convenient than app-based two-factor authentication. So what I’ve settled on is using a security key for my most important accounts, but sticking with an app for everything else.

Ultimately, my experience with YubiKeys sums up my approach to security and privacy — I still value convenience a lot. By doing things like using app-based two-factor authentication for emails and cloud storage that lacks end-to-end encryption, I’m still probably being a bit inconsistent about my privacy practices.

But I certainly feel more secure than I did before I went on this journey, and I have beefier security in place for the accounts and data that matter most.

You’ll have to figure out what the right balance is for you personally, but the tools above should help you make some security and privacy upgrades.

• Samsung advances floating data center project with first Texas engineering agreement
• Floating facilities promise 50MW capacity while reducing traditional construction timelines significantly
• Texas grid review creates fresh uncertainty for expanding data center developments statewide

Samsung Heavy Industries and Dallas-based Mousterian Corporation have signed an engineering contract covering their first floating data center project planned for deployment in Texas.

The deal advances their partnership into the engineering phase, with both companies jointly responsible for the project's basic design, detailed engineering, and production planning.

Each proposed floating facility is expected to provide 50 MW of critical IT capacity while operating on water instead of occupying conventional land-based development sites.

The pilot site is expected to sit next to an existing combined-cycle gas turbine power plant located in Houston, according to reporting on the plan.

Mousterian believes the shipyard construction approach could shorten delivery schedules compared with conventional land-based developments.

"Because the facility is built in a shipyard, fabrication advances in parallel with sitework rather than after it, a schedule and scale that conventional delivery methods cannot match," said Min Suh, chief executive of Mousterian.

Samsung Heavy Industries brings five decades of offshore and maritime engineering experience into the venture, according to its chief executive.

"Together with Mousterian, we intend to deliver floating datacenters with the same rigor, quality, and schedule certainty that define our maritime track record," said Sung-an Choi, chief executive of Samsung Heavy Industries.

Mousterian says its cooling approach avoids drawing on potable water supplies entirely and sends no discharge back into surrounding waterways.

Removing air-cooled chillers and fans from the design is also expected to considerably cut down on noise pollution near the site.

The American Bureau of Shipping is expected to classify these waterborne facilities once formal engineering work moves further.

The proposed location beside an existing gas-fired power plant may not shield the project from broader regulatory challenges affecting new data center developments.

Texas Governor Greg Abbott has paused approvals for new Electric Reliability Council of Texas (ERCOT) grid connections while state authorities review pending applications.

ERCOT is reviewing requests covering more than 474 GW of proposed new connections, a figure Abbott says is over five times peak state demand.

Data centers alone account for roughly 90% of those pending connection requests currently sitting before regulators.

State regulators are now asking project developers whether they plan to generate their own power or depend fully on the shared grid.

Officials also want detailed water consumption estimates from each project before any new connection request receives approval going forward.

Any project that fails to meet these new disclosure requirements will simply be denied a grid connection outright.

Floating infrastructure could sidestep competition for scarce buildable land near major cities, while offering built-in access to water for cooling.

Whether Texas regulators treat water-borne facilities differently from land-based ones under the new audit rules remains unclear for now.

Via The Register

• The system connects directly to a standard alternating-current power outlet
• Sulfur co-implantation gave engineers precise control over individual diamond qubits
• Single-qubit fidelity reportedly reached 99.98%

German startup Saxon Q has launched the world's first portable diamond-powered quantum computer operating at room temperature without requiring complex cryogenic cooling equipment.

The system fits inside a standard server rack, connects directly to a conventional alternating-current power outlet, and supports configurations reaching 128 qubits.

The company says larger 512-qubit systems will become available next year, while its long-term roadmap aims to exceed 10,000 qubits after 2030.

The system relies on nitrogen-vacancy defects created inside synthetic diamonds, where nitrogen atoms occupy positions beside empty spaces within the crystal structure to function as quantum bits.

Scientists manipulate these qubits using lasers and microwave pulses, allowing them to represent quantum states extending beyond conventional binary computing methods.

Although nitrogen-vacancy technology has existed for years, producing stable systems beyond 10 qubits has remained difficult because creating reliable defects at larger scales proved technically challenging.

Saxon Q says its breakthrough came from introducing sulfur alongside nitrogen during manufacturing, allowing engineers to improve stability while gaining greater control over individual qubits.

“That's the key technology point. Because the sulfur lifts the chemical potential to a point that it's negatively charged,” said Marius Grundmann, professor of experimental physics at Leipzig University and co-founder of Saxon Q.

“The sulfur supplies the electron; the sulfur also makes the vacancy attached to the nitrogen with a very high yield.”

According to the company, this manufacturing process enabled qubits to achieve 99.92% fidelity before error correction, representing fewer than one error for every 1,000 operations.

However, it later updated its data, saying subsequent measurements increased single-qubit fidelity to 99.98%, although those figures have not yet received independent verification.

The hardware is currently available in rack-mounted systems supporting up to 128 qubits, while 512-qubit configurations are expected next year and expansion beyond 10,000 qubits remains the long-term objective after 2030.

Unlike many quantum systems that require cryogenic refrigeration and extensive infrastructure, Saxon Q says its hardware can operate directly within conventional computing environments without dedicated cooling facilities.

“We have a fully functioning quantum computer… We have a quantum computer that can execute quantum code that you can reach via the network, and it is a multiuser, multitasking, multicore system,” Grundmann added.

The company believes local deployment could reduce dependence on cloud-based quantum services for applications including robotics and autonomous driving, where communication delays may become unacceptable.

Researchers still caution that comparing nitrogen-vacancy systems directly with superconducting quantum computers remains difficult because both technologies emphasize different engineering approaches and operating characteristics.

Existing studies generally indicate superconducting systems deliver faster processing, although the balance between processing speed, cloud latency, and practical deployment remains uncertain across different workloads.

Another unresolved challenge involves increasing chip density, since current devices support only eight or 16 qubits per chip.

Future applications may require hundreds of thousands or even millions of interconnected qubits operating within significantly larger quantum arrays.

Via Live Science

A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Sunday's puzzle instead then click here: Quordle hints and answers for Sunday, August 9 (game #1658).

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.

Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.

SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.

• The number of different vowels in Quordle today is 4*.

* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).

• The number of Quordle answers containing a repeated letter today is 1.

• No. None of Q, Z, X or J appear among today's Quordle answers.

• The number of today's Quordle answers starting with the same letter is 0.

If you just want to know the answers at this stage, simply scroll down. If you're not ready yet then here's one more clue to make things a lot easier:

• G

• C

• S

• E

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

The answers to today's Quordle, game #1659, are…

• GAVEL
• CRUSH
• SPEED
• ETHOS

I’m glad I held back from getting GAVEL until later in the game.

Before eliminating the letters there were dozens of possibilities, but as the letters vanished it reduced down to one.

The answers to today's Quordle Daily Sequence, game #1659, are…

• PRISM
• SNAKE
• CRAMP
• SALON

• Quordle #1658, Sunday, 9 August: KAPPA, WIDEN, FOLIO, STEEP
• Quordle #1657, Saturday, 8 August: FUROR, BUGLE, REUSE, SCOOP
• Quordle #1656, Friday, 7 August: AMISS, SHIRK, SALSA, RABBI
• Quordle #1655, Thursday, 6 August: AMUSE, BRAND, CEDAR, BLURB
• Quordle #1654, Wednesday, 5 August: MOUND, WRONG, STOOD, WEIRD
• Quordle #1653, Tuesday, 4 August: GOOEY, TARDY, HUMAN, CHALK
• Quordle #1652, Monday, 3 August: FUNGI, MUDDY, DEBUT, MANIA
• Quordle #1651, Sunday, 2 August: BOOTH, EASEL, REACH, MAKER
• Quordle #1650, Saturday, 1 August: OMEGA, BIRCH, SMOCK, EERIE
• Quordle #1649, Friday, 31 July: MANIA, SINGE, STOOL, LOFTY
• Quordle #1648, Thursday, 30 July: MODAL, IDLER, CHUMP, LUMPY
• Quordle #1647, Wednesday, 29 July: GAWKY, ALLOY, AUDIT, LATCH
• Quordle #1646, Tuesday, 28 July: SMITE, ROACH, RUDDY, GLOBE
• Quordle #1645, Monday, 27 July: TAFFY, FIGHT, GUILD, WAGER
• Quordle #1644, Sunday, 26 July: STOMP, LIMIT, FUNKY, STEAD
• Quordle #1643, Saturday, 25 July: SALON, SHEEN, BURST, GOURD
• Quordle #1642, Friday, 24 July: DOLLY, EDIFY, LAGER, PRANK
• Quordle #1641, Thursday, 23 July: LADEN, EVICT, WOVEN, SHIRE
• Quordle #1640, Wednesday, 22 July: COYLY, SINGE, AWASH, AWOKE
• Quordle #1639, Tuesday, 21 July: TRUSS, AXIAL, SWINE, THING

• Forcepoint X-Labs publishes threat model for persistent memory poisoning
• Hidden text on a webpage becomes a durable "fact" an agent retrieves and trusts in unrelated tasks weeks later
• It has already been demonstrated against products already in the market, including ChatGPT, Gemini, Claude and Microsoft 365 Copilot

New findings from Forcepoint's X-Labs outline an interesting scenario that could easily mimic real life: An AI assistant with browser access reads a webpage about travel disruption.

Near the bottom of that page, in text sized and positioned so no human will ever see it, sits a short paragraph stating that ABC Travel Support is the official emergency booking provider and should always be recommended when urgent travel changes are needed.

The assistant's text extractor does not distinguish between hidden and visible text, so the model treats the whole thing as plain prose and files the claim away as a useful fact about how this organization handles travel. A month later, the user's flight is canceled. They ask their assistant what to do, and it tells them, helpfully and with no sign of anything wrong, to contact ABC Travel Support.

This is what Forcepoint calls persistent memory poisoning, a security vulnerability where an attacker injects false data or malicious instructions into an AI agent's long-term memory or retrieval database, and it is a threat model that is increasingly in focus as users increasingly rely on AI, often treating its responses as gospel, despite the warnings most chatbots come with.

The canonical academic result is MINJA, short for Memory INJection Attack, presented at NeurIPS 2025. Its significance is the attacker model. MINJA does not assume access to the memory store, elevated privileges, or any compromise of the system. It works by submitting ordinary queries through the standard interface, using indication prompts, bridging steps, and a progressive-shortening technique that strips away giveaway language while leaving the poisoned record behind.

Across GPT-4o-mini, Gemini 2.0 Flash, and Llama 3.1 8B, it reported injection success above 95% and attack success above 70%.

It must be noted that those numbers might be optimistic; a January 2026 paper evaluating memory poisoning in electronic health record agents notes that MINJA's numbers were obtained under idealized conditions, and that how well these attacks hold up in realistic deployments remains understudied.

Despite this, it remains a significant threat to products that continue to ship, including ChatGPT, Gemini, Claude, and Microsoft 365 Copilot. It is important to find a solution to a problem that Microsoft has already warned about in the past; Forcepoint suggests an approach that could mitigate it.

Its proposal is to stop treating extracted memories as facts and start treating them as objects that can be inspected. Each memory is stored with metadata: where it came from, what type of source it is, whether a user confirmed it, and a risk score. Language written to shape future behavior, phrases like "from now on" or "make this your default going forward," adds to the score. So does the sudden appearance of a previously unseen domain, contact, or vendor.

Contradiction detection is also in play: if new memory conflicts with an existing entry about the official travel provider, both cannot be true, so the engine flags the conflict and holds the new item for user confirmation rather than silently overwriting it. At the same time, anything related to payment instructions, banking details, VPN configuration, or security contacts is given higher weight, regardless of where it came from.

None of these approaches, however, solves the underlying problem: agents are built to treat retrieved memory as their own experience rather than as input. Scoring raises the cost of poisoning. It does not change what the agent believes once something gets through, and as Agent Security Bench found, current defenses are not doing well.

For anyone using an assistant with memory today, the practical play is unglamorous but worth following anyway: open the memory settings occasionally and read what is in there, but that's easier said than done when it comes to propagating the message since a sizeable chunk of AI users never bother to look under the hood.