News
- Accomplish AI showed Claude Cowork could escape a VM sandbox via Linux zero‑day CVE‑2026‑46331
- Agent accessed host Mac files, risking exfiltration of SSH keys, cloud credentials, and more
- Anthropic shifted Cowork to default cloud execution; local users must harden configs to mitigate exposure
Recent news of a ChatGPT agent escaping the sandbox and attacking services on the internet raised quite a few eyebrows, but it seems it’s not the only one capable of running wild. Security researchers Accomplish AI are saying they achieved similar results with Anthropic’s Claude Cowork.
In a new report, the researchers said they ran a local session in a Mac-hosted virtual Linux machine and then observed as the agent broke free of the VM and started reading and writing files on the underlying system.
“We connected a folder to a fresh Claude Cowork session, sent one short message, and watched the agent escape the sandbox,” Oren Yomtov, principal security researcher at Accomplish AI, told The Hacker News. “From inside the VM, it reached the host Mac and read and wrote files all over it, far outside the folder we'd connected, with no permission prompt anywhere.”
Defaulting to cloud executionThis means that, in theory, the agent can be used to access or exfiltrate anything that’s stored on the Mac’s user account, including SSH keys, cloud credentials, and more. To break out of the sandbox, the agent exploited CVE-2026-46331 ("pedit COW"), a Linux kernel privilege-escalation vulnerability. This flaw, fixed in mid-June this year, was given a severity score of 7.8/10 (high).
Accomplish AI disclosed these findings with Anthropic, which allegedly acknowledged them but did not issue a direct fix. However, the version of Claude Cowork that was released afterwards defaults to cloud execution which, the publication claims, addresses the issue. Still, users who opt to run the agent locally rather than in the cloud will remain exposed.
Mitigations are possible, though. Users should disable unprivileged user namespaces, grant/revoke seccopm permissions, stop modules autoloading, and restrict sharing of the whole host into the VM.
"Scope it to the folders that were actually connected instead of all of /, or at least mount it read-only, and run coworkd with ProtectSystem=strict in its own mount namespace so it isn't re-execing binaries a session user can poison," Accomplish AI explained. "Then even a full guest-root has nothing to land on, the last two steps of the chain have nowhere to go."
- Stream Shark Week 2026 free via YouTube TV / DirecTV / Hulu + Live TV (US)
- Unlock your stream with Surfshark (save up to 85%)
- Shark Week 2026 starts at 8pm ET/PT on Sunday, July 26
"It's all psychological. You yell barracuda, everybody says, "Huh? What?" You yell shark, we've got a panic on our hands." We're still waiting on Barracuda Week (which we'd definitely tune into), but in the meantime TV event of the summer Shark Week is back to keep you glued to your telly and out of the Amity Island waters.
Anything to keep Mayor Vaughn happy.
Even better – you can stream all 20 specials across seven nights, including Air Jaws: Red, White and Breach, How to Train a Great White and My Strange Shark Addiction, for free.
Here's the trick: Shark Week airs exclusively on Discovery, which you can access with a free trial, and Surfshark will let you watch Shark Week from anywhere. Not only is the aptly named VPN excellent, but it's also heavily discounted for a limited time, and comes with a 30-day money-back guarantee.
How to watch Shark Week 2026 for freeDiscovery is home to Shark Week 2026. If you don't have the channel on cable, it's available on a streaming basis from several cable alternatives. Best of all, a few of them let new subscribers take advantage of a free trial.
YouTube TV usually costs $82.99 per month but comes with a 5-day free trial and $15 off each of your first five months.
The DirecTV MyEntertainment genre pack costs $42.99 per month after a 5-day free trial, and Hulu + Live TV costs $89.99 per month after a 3-day free trial.
Outside the US? You can use a VPN to watch Shark Week 2026 for free as if you were right at home.
Use a VPN to watch Shark Week 2026 for free from anywhereSurfshark CHEAPEST Deal – from £1.79/month
Surfshark is our favourite cheap VPN – and it's on sale today! We use it to watch streaming services from abroad, block ads, protect from viruses, and more.
✅ Up to 85% Off Today
✅ 3 Extra Months Free
✅ 30-day money-back guarantee
✅ Unlocks all necessary streaming services
Get Surfshark today and stream Shark Week 2026 free from anywhere. View Deal
Quick start: Using a VPN to watch Shark Week 2026 freeUsing a VPN is incredibly simple; just follow these steps.
1. Install the VPN of your choice. As we've said, Surfshark is the one we recommend.
2. Choose the location you wish to connect to in the VPN app. For instance if you're visiting the UK and want to view an American service, you'd select the "United States" location from the server list.
3. Sit back and enjoy the action. Head to your streaming platform of choice and watch Shark Week 2026 just like you would at home.
What to know about Shark Week 2026(All times ET/PT)
Sunday, July 26
8pm – K-Pop Shark Heroes
9pm – Air Jaws: Red, White and Breach
10pm – Invasion of the Mega Sharks
Monday, July 27
8pm – Bull Shark Dinner Bell
9pm – House of Sharks
10pm – Biggest Mako on Earth
Sunday, July 26
8pm – Jurassic Sharks
9pm – Jaws vs Orca
10pm – Chum Island: Catching a Killer
Sunday, July 26
8pm – Expedition X: Atomic Sharks
9pm – Expedition Unknown: Shark Secrets
10pm – Alien Sharks: Untamed America
Sunday, July 26
8pm – What Shark Attacked?
9pm – How to Train a Great White
10pm – Ultimate Shark Dive
Sunday, July 26
8pm – Secrets of the Great White Kill
9pm – Sharkzilla Takes New York
10pm – My Strange Shark Addiction
Sunday, July 26
8pm – Great White Highway
9pm – Thresher Shark: Stun to Kill
We test and review VPN services in the context of legal recreational uses. For example:1. Accessing a service from another country (subject to the terms and conditions of that service).2. Protecting your online security and strengthening your online privacy when abroad.We do not support or condone the illegal or malicious use of VPN services. Consuming pirated content that is paid-for is neither endorsed nor approved by Future Publishing.
The GameSir G8+ Bluetooth is the wireless mobile controller in the G8 family and bolsters compatibility from being simply mobile, to the Nintendo Switch, tablets, and PC. The same full-size grips, Hall effect joysticks and triggers, and programmable back buttons that made the wired versions of the G8 so impressive remain, and at $79.99 / £79.99 / AU$129, so does the price.
On paper, you’d think it’s a no-brainer to go wireless, particularly with the addition of a six-axis gyroscope adding to that strong core experience, but the wireless freedom you gain comes with meaningful trade-offs. Noticeably, the lack of passthrough charging and the lack of a headphone jack, both obvious casualties of the move to Bluetooth.
More importantly for iPhone gamers, the GameSir iOS app doesn't recognize the controller, which means the stick calibration, dead zone tuning, hair-trigger configuration, and back button mapping that make the G8 Plus so compelling as an option on other platforms go completely to waste. The Bluetooth connection itself also feels less locked-in than the wired experience on iPhone, with small stick movements occasionally getting lost along the way.
On Android and PC, the experience is significantly better. The app works, the back buttons are configurable, and there’s haptic feedback on offer as well (albeit with patchy support, particularly on mobile). For the first time, you can even use the G8 Plus with your Nintendo Switch. But if an iPhone is your primary device, the wired GameSir G8+ MFi delivers a more complete, more reliable experience for the same money.
(Image credit: Future)GameSir G8+ Bluetooth: Price and availability- List price: $79.99 / £79.99 / AU$129
- Available from GameSir and major retailers
- Same price as the rest of the G8 range
The middle sibling, the GameSir G8+ Bluetooth launched in mid-2024 at $79.99 / £79.99 / AU$129, matching the rest of the G8 family.
Each variant has pros and cons, so the consistent price puts the onus firmly on the buyer to decide which set of trade-offs works best for their setup. GameSir has also released a couple of special edition color variants in some regions, including a Hulk and a Thanos Edition, which will set you back slightly more.
At this price, it remains highly competitive against the Backbone One 2nd Gen and is significantly cheaper than the Razer Kishi Ultra, though the Bluetooth limitations on iPhone mean it's not the straightforward recommendation the wired models are.
GameSir G8+ Bluetooth: SpecsGameSir G8+ Bluetooth
Price
$79.99 / £79.99 / AU$129
Dimensions
9.06 x 4.21 x 2.17 in / 230 x 107 x 55mm
Weight
0.69 lbs / 314g
Connection
Bluetooth 5.3, Wired USB-C (PC), Wireless Dongle (sold separately)
Compatibility
iOS, Android, Nintendo Switch, PC, macOS
Software
GameSir App (Android only)
GameSir G8+ Bluetooth: Design and features- Comfortable, console controller-like ergonomics shared with the rest of the G8 range
- Bluetooth versatility across five platforms, but mode switching is a hassle
- Special edition designs are only available in certain markets
The G8+ Bluetooth is the first and only member of the G8 family that cuts the cord entirely. Wireless freedom and the versatility it brings are the entire pitch, and it helps broaden the G8’s platform compatibility too.
Unlike the original G8 Galileo and the G8+ MFi, which are limited to mobile only, this controller works with iOS, Android, the Nintendo Switch, PC (via wired USB-C or Bluetooth, with a wireless dongle sold separately), and macOS. Console gaming aside, in theory, that means the G8+ is ready to follow you anywhere on your gaming journey.
The reality, unfortunately, is a little more complicated. Each platform requires its own Bluetooth pairing mode, triggered by holding different button combinations on the controller. If you're a one-device person, this is a simple pair-once-and-forget situation, and you can skip to the next paragraph.
If you're hopping between platforms, as I was during testing, it gets tedious fast. Switching modes caused regular reconnection failures with devices the controller had previously been paired with, and keeping track of which mode goes with which platform isn't intuitive. It's workable, and we’re in first-world problem territory here for sure, but it's not the seamless experience it should be.
FutureFutureOn the plus side, turning it on couldn’t be simpler because stretching the controller open automatically powers it on and letting it close fully again powers it down. This isn’t a life-changing feature but certainly a nice convenience that worked reliably throughout testing.
A 1000mAh battery sits split across both grips, charging via USB-C. The GameSir app tracks each side independently, which is how I discovered the left and right grips drain at slightly different rates despite charging as a single unit. I saw a couple of percent drop over half an hour of active use, and the rated eight-hour battery life seems about right based on that.
The black colorway is a clean, classic look that feels more like a traditional controller than the grey and grey-and-white of its wired siblings. GameSir has also released rather jazzy Hulk- and Thanos-themed special editions in some regions, making this the only G8 variant with cosmetic variety, though they’ll set you back a little more for the paint job.
The build quality and overall shape match the rest of the family; it’s well designed and well built with the same full-size, Xbox-like grips, textured backs and a solid chassis. It’s comfortable to hold for extended sessions, and while at 11oz/314g it's the heaviest of the three on paper, the balance is good, and the weight didn't register during testing.
Case compatibility is a mixed bag, as is often the case with mobile controller grips. Without a USB-C connector going into the phone, there's no port angle to stress about, thankfully, but the physical cradle can still be tight.
My iPhone 17 Pro fit with a slim case on but sat at a noticeable angle towards the left side, which saw it protrude more than I’d like. There is a rubber pad on the phone deck that appears to serve the same purpose as the G8+ MFi's removable magnetic panel for camera bump clearance, but this one is glued down rather than magnetically attached.
GameSir’s marketing suggests this is still meant to be removed to accommodate camera bumps, but it really didn't feel like it wanted to come off, and I wasn't willing to force it. If you do take it off, don’t expect it to go back on neatly and expect to trap plenty of dust in the residue.
The D-pad shares the same feel as the rest of the G8 family: slightly mushy initial movement but with a clicky, short-travel feel once you commit to an input. Swappable ABXY caps and thumbstick caps carry over, too, as do the three alternate stick options, and they’re easily adjusted by popping off the magnetic faceplates.
Once again, however, GameSir only includes one of each of these replaceable sticks, so you can’t run the same option on both sides. One annoyance that cropped up here for the first time on any G8 I’ve tested: on iOS, the controller connects as a DualShock controller, so all on-screen UI elements in compatible games display PlayStation button labels.
GameSir doesn't include any PlayStation-legended alternative keycaps in the box, so you’ll need to remember that triangle is Y, for example.
(Image credit: Future)GameSir G8+ Bluetooth: Performance- Hall effect inputs are excellent and responsive, but Bluetooth feels a step behind wired
- iPhone users lose app support, back buttons, and customization entirely
- Vibration motors remain largely a spec-sheet feature with no real-world use
As you’d expect, the core gameplay experience with the G8+ Bluetooth mirrors the other options in the G8 range. The Hall effect thumbsticks and triggers are smooth and precise with a high level of sensitivity.
Resistance in these is just right for my taste; I found I was able to confidently hold the sticks at a midpoint for fine adjustments without the sense that the stick was itching to return to its starting point. The triggers offer full analog travel with a responsive hair-trigger mode (assuming you can enable it, more on that shortly). In analog mode, I felt like I had a nice level of control and was able to modulate the throttle in Forza Horizon 6 with plenty of detail.
That said, the Bluetooth connection introduced a subtle but noticeable difference in feel I hadn’t experienced before with a GameSir G8. I couldn’t quite quantify it, but I didn't feel as locked-in using the G8+ Wireless as I did with the wired equivalents.
It's not consistent, obvious latency, or any kind of genuine stutter or interference; it's more that small thumbstick movements occasionally felt like they were getting lost along the way. Everything was perfectly playable, and my skill as a gamer is certainly not high enough for this to be a make-or-break thing, but if you're serious about your gaming, the wired experience is definitely tighter.
(Image credit: Future)The biggest issue I came across, and the one that separates this controller from the rest of the G8 family, is the iPhone experience. It’s somewhat lacking. Because iOS sees the controller as a DualShock 4, the GameSir app doesn’t recognize the G8+ Bluetooth. That means no stick calibration, no deadzone adjustment, no hair-trigger toggling, and no back button mapping.
The two rear programmable buttons, labeled M1 and M2, ship unbound by default, and there's no way to configure them on iOS. They don't appear in the iPhone's native controller settings either because iOS thinks this is a DualShock 4, which doesn't have back buttons. On macOS, I was able to rebind other buttons through native system options, but back button mapping wasn't available there either for the same reason.
On Android, the picture is much better. The app recognized the controller (though admittedly it took a couple of attempts at first), and once connected, all the usual calibration, deadzone, hair-trigger, and back button configuration options were available. Gaming on the POCO X5 Pro was a smooth experience, with Fortnite and Call of Duty Mobile both running well after the usual game-side controller recognition quirks.
PC gaming was the most pleasant surprise. Connected via either USB-C or Bluetooth, the controller was recognized immediately as a PlayStation 4 controller and was fully mapped out of the box for both local and cloud gaming. Gyro tracking worked here too, which is a nice bonus for games that support it. On macOS, things were less smooth.
The controller initially tried to connect as a "GameSir G8" which required manual Steam configuration and then wasn't recognized in-game. Reverting to DualShock mode fixed the issue, and games worked with automated mapping after that.
Shoudl you buy the GameSir G8+ Bluetooth?Buy it if...You're an Android user who wants one controller for phone, Switch, and PC
The Bluetooth versatility and full GameSir app support on Android make the G8+ Bluetooth a seriously flexible, wireless option. If you play across multiple platforms and want a single controller to handle them all, this could well be it.
You value wireless convenience over raw precision
No connections mean a simpler experience and fewer points of failure. It also brings a wider variety of devices into play.
You're primarily an iPhone user
While it does work natively for games that play nicely with a physical controller, the lack of GameSir app support, non-functional back buttons, and DualShock workaround make the iPhone experience feel like an afterthought. The wired G8+ MFi is the better choice for the same money and with little drawback.
You want the tightest possible input response
Don’t get me wrong, the Bluetooth connection is perfectly playable, but it doesn't feel as precise and instantly responsive as the wired G8 controllers. If competitive precision matters to you, go wired.
Trying to work out if the GameSir G8+ Bluetooth is the one to get? Here's how it compares to other, similar controller grips.
GameSir G8+ Bluetooth
GameSir G8+ MFi Galileo MFi
Scuf Nomad
Price
$79.99 / £79.99 / AU$129.99
$79.99 / £79.99 / AU$129.99
$99.99 / £89.99 / around AU$179.99
Dimensions
9.06 x 4.21 x 2.17in / 230 x 107 x 55mm
9.02 x 4.20 x 2.13in / 229 x 106.8 x 54.2mm
6.93 x 3.70 x 1.34in / 176 x 93.98 x 34.03mm
Weight
10.72oz / 314g
10.72oz / 304g
0.5 lbs / 228 g
Connection
Bluetooth 5.3, Wired (USB-C), Dongle (sold separately)
Wired (USB-C)
Wireless (Bluetooth)
Compatibility
Android, iOS, Nintendo Switch, PC, macOS
Android, iOS
iOS
Software
GameSir App (Android only)
GameSir App
Scuf Nomad Companion App (iOS)
GameSir G8+ MFi
The G8+ MFi is the wired, Apple-friendly equivalent from the same family. It offers MFi certification, full GameSir app support on both iOS and Android, passthrough charging, a headphone jack, and functional back buttons on every platform. If you don't need wireless or Switch/PC compatibility, it's the more complete package.
For more information, check out our full GameSir G8+ MFi review
Scuf Nomad
The Scuf Nomad is an iPhone-only Bluetooth alternative with a lighter build, 16 hours of battery life, and an app that actually works on iOS. It's more compact, handles phone cases better, and has configurable rear paddles. The trade-off is no Android, Switch, or PC support and a higher price tag.
For more information, check out our full Scuf Nomad review
How I tested the GameSir G8+ Bluetooth- Tested with multiple devices
- A wide variety of games tried
- Compared to other mobile controllers
I spent several days testing across iOS, Android, PC, and macOS, with a focus on comparing the experience to the wired G8 and G8+ MFi I'd previously reviewed.
I tested on an iPhone 17 Pro and POCO X5 Pro 5G, playing Fortnite, Call of Duty Mobile, and Forza Horizon 6 via Xbox Cloud Gaming on mobile, and Stardew Valley via Steam on PC. I evaluated Bluetooth connectivity, app compatibility, back button functionality, latency feel, vibration motors, and cross-platform mode switching.
First reviewed June-July 2026
- Cloudflare finds kick-off time, not match importance, drove the biggest World Cup traffic swings
- Overnight matches more than doubled normal traffic in some countries, while daytime fixtures barely registered
- The tournament's biggest internet moment ironically wasn't the final but Argentina's 3-1 quarter-final win over Switzerland, which moved global traffic by a factor of 1.26x.
Although Spain's 1-0 extra-time win over Argentina closed out the FIFA 2026 World Cup, the tournament's biggest traffic pulls actually came days before.
Cloudflare has published its own analytics, tracking the likes of HTTP traffic, DNS, and security trends across a network spanning more than 330 points of presence in over 120 countries across June and July 2026.
While the analytics confirm some widely held assumptions about the World Cup, it also demolishes others, as it found it was not the final which moved the internet most, but a quarter-final involving Argentina.
A World Cup with plenty of interesting insightsPer Cloudflare's data, the Argentina vs Switzerland quarter-final topped every other fixture with a deviation factor of about 1.26, while the final ranked eighth.
Somewhat unsurprisingly, Argentina, the defending champions, also moved the most traffic during their run, at 1.18x that of a typical country, followed by France at 1.16x.
Matches played between roughly midnight and 8 am local time produced the biggest deviations, in some cases more than doubling normal traffic. The reason is straightforward: those are hours when almost nobody is online, so fans staying up or waking early to watch represent a large proportional swing.
Matches falling between 9am and mid-afternoon barely registered, since the people watching would have been online regardless. Early-evening kickoffs produced a smaller secondary lift, most visible on weekdays, as matches kept people connected during hours when usage would normally taper off.
(Image credit: Marvin Ibo Guengoer - GES Sportfoto/Getty Images)Bosnia and Herzegovina illustrated both ends of this. When the national team played at 2 am local time, traffic at points more than doubled. When they played in the evening, traffic fell to around 70% of typical levels as people put their devices down and simply watched.
The clearest demonstration came from Brazil's 2-1 Round of 32 win over Japan on June 29. The same 90 minutes fell during ordinary waking hours in Brazil and in the dead of night in Tokyo. Japanese traffic ran at roughly double normal levels; Brazilian traffic sat about a quarter below normal. Two nearly parallel curves, opposite sides of the baseline, same match.
Grouping countries by the shape of their match-day traffic curves produced three clusters, and the split reveals something about how people actually watch.
The largest group (44 countries across 101 matches) saw traffic rise during halftime and hydration breaks, the familiar pattern of everyone reaching for their phone the moment play stops. A smaller group of eight countries across 18 matches did the exact opposite, with traffic falling during those same pauses.
The explanation turned out to be streaming. That second cluster (Algeria, Tunisia, Jordan, Egypt and DR Congo) showed a traffic mix dominated by multimedia and streaming requests, meaning a large share of viewers were watching the match online rather than on broadcast television.
Notably, hydration breaks had little effect in the streaming cluster but a clear one elsewhere. Cloudflare's reading is that a viewer won't close a stream for a three-minute cooling break but will for a fifteen-minute interval. In the majority of countries, though, even those three minutes were enough to produce a visible traffic bump: Croatia and Bosnia and Herzegovina showed the effect most strongly.
Requests to gambling industry sites rose overall from the opening game onward. The more interesting change was structural. Before the tournament, betting traffic followed a pronounced weekly rhythm; afterward it flattened into a near-constant profile, a consequence, Cloudflare suggests, of matches running at near-daily regularity for a month.
At the scale Cloudflare operates, it is genuinely difficult for any single event to leave a visible mark on aggregate global traffic. That the 2026 World Cup managed to do so is the clearest measure of how much the tournament meant to people online.
The contentious issue of data centers — and their impact on the environment and local communities — continues to make headlines, as the need for more computing power comes up against concerns about whether that extra capacity is worth the cost.
One of the natural resources that data centers need is water, and plenty of it, to cool down the chips running calculations. For regions already living through rising temperatures and climate change, that's a big ask, and it's understandably getting a lot of attention.
It's also something the tech giants building these data centers are aware of, and Google has just announced a strategy for using water more responsibly: "how we build is just as important as what we build," Google says. Here's what the plans are, and the size of the problem we're dealing with.
Data center water usageData centers are nothing new of course, and are needed for everything from hosting this TechRadar website to delivering movies on Netflix. With the rise of AI, however, there's been a need for more data center capacity, and a higher level of performance from data centers — which all adds to the amount of water needed for cooling.
If you find data center water usage figures confusing, you're certainly not alone. There's a lot of variation in the statistics used to measure it, for all kinds of reasons: data centers water usage can vary significantly from site to site, for example, and there's also the question of what counts as water usage.
Should the water needed to generate the electricity to power a data center be included, or not? What standard is used to measure water 'lost' to the environment if it's returned to the local water cycle? These questions are answered differently in different reports, hence the discrepancies in figures — and the problem is compounded by the fact that many data centers simply don't report their water usage at all.
Data center water usage is in the spotlight (Image credit: May Cloud/Unsplash)While these figures are a little out of date now, a 2021 study put direct (on-site) data center water usage at 1.7 billion liters per day in the US, compared to a national water consumption of 1,218 billion liters per day — so that's less than 1%. As is often pointed out, golf course usage is about four times more.
A few points worth noting: that doesn't account for indirect water usage, and data center demand is rising fast (unlike, say, water for golf courses). Most importantly though, the significant impacts of data center installations are felt at the local level, not the national level. The largest data centers can use up the same amount of water as a town of 50,000 people, which has clear implications for the people in that town.
Many data centers are being built in water-stressed areas, and then there are the other factors to consider, like electricity demand and noise pollution. While water usage for data centers at a national level may be a small fraction, that doesn't mean it isn't a major problem at a local level, and the problem is getting worse.
What Google is proposingThere are five commitments that Google is making. First, more water will be added back into the system than is consumed by Google data centers by 2030, mostly by separate projects looking at more efficient water use and reduced waste (a little bit like planting trees to offset plane trips).
Second, the company is going to invest hundreds of millions of dollars into water infrastructure, to make sure areas where data centers are located are kept well supplied. Third, Google says it will use air-cooled or recycled water data centers in places where water is particularly scarce (these solutions use more electricity and are more expensive, but they don't demand as much water).
Fourth, Google is committing to being transparent when it comes to the water use of its data centers — something that is by no means standard across the industry at the moment. Fifthly and finally, Google says it will "pursue alternative and reclaimed solutions to protect water resources", including looking at reclaimed wastewater.
Google is using water stewardship projects to offset data center demands (Image credit: Google)"We think it’s really important to sort of put a blueprint out there that communities can reference," Google's Ben Townsend told The Verge. "So if somebody else comes and says, 'we'd like to build a data center there', a community can say, 'well, here are five different things that really put the community and the watershed first... are you doing these?'"
Water usage is only part of the data center story, and it's a story that's becoming increasingly complex and multi-faceted. But positive action (or at least the promise of positive action) is better than nothing, and this may help to shift the industry towards more responsible and sustainable infrastructure.
Turning around public sentiment might be more difficult than cutting down on water usage though. While the hype from AI companies continues, and huge changes are being seen in areas such as software coding, there are still a significant number of people who aren't keen on generative AI and what it brings with it — even before getting to the issue of how many natural resources it uses up.