News

Back in June, Apple raised the price of its iPads due to "significantly higher memory costs," but Amazon Prime Day has at least made these excellent tablets a little more affordable in the short term.

• Browse the full Amazon Prime Day sale

In the US, for instance, the iPad mini 8.3-inch (2024) is now $499 (was $599), which is equal to its starting price pre-RAM crisis. Frustratingly, the same tablet has been available for as little as $399 in previous Prime Day sales, but with the RAM crisis expected to last at least another three years, the chances of ever seeing such low iPad prices again are extremely slim.

Meanwhile, in the UK, the iPad Air 11-inch (2026) is currently on sale for £669 (was £749). That's not quite as low as its pre-rise £599 price, but it's not far off. Beggars can't be choosers, as the saying goes.

Here's a roundup of all the other iPad deals currently available at Amazon in both the US and UK:

• iPad 11-inch (2025): was $449 now $399
• iPad mini 8.3-inch (2024): was $599 now $499
• iPad Air 11-inch (2026): was $749 now $649
• iPad Air 13-inch (2026): was $949 now $849
• iPad Pro 11-inch (2024): now $2,099 now $1,899
• iPad Pro 13-inch (2024): was $1,499 now $1,299
• iPad Pro 11-inch (2025): now $1,199 now $1,099.99
• iPad Pro 13-inch (2025): was $1,499 now $1,399.99

• iPad 11-inch (2025): was £429 now £409.97
• iPad Air 11-inch (2026): was £749 now £669
• iPad Air 13-inch (2026): was £949 now £869
• iPad Pro 11-inch (2025): was £1,199 now £1,159

As for which iPad you should choose, well, that depends on your specific usage habits. We’ve got a dedicated guide to the best iPads elsewhere on TechRadar, but I'll give you a quick rundown of each model's key features here.

The iPad 11-inch (2025) is the model to choose if you want a tablet for streaming movies, scrolling social media, and FaceTiming family and friends. It's not a workhorse by any means, but with an 11-inch LED display and a capable A16 chipset, Apple's most basic iPad is still the best tablet you can get for around the $400 / £400 mark.

The smaller iPad mini 8.3-inch (2024) bumps up the power with a more powerful chipset and Apple Intelligence compatibility, but honestly, you probably won't notice these features unless you're planning on using your iPad as a productivity tool, in which case I'd recommend jumping straight to the iPad Air instead.

Why? Well, firstly, the iPad Air 13-inch (2026) is available in two more work-friendly sizes than the iPad mini. It's also got a MacBook-level M4 chipset and 12GB of RAM, so it can power through productivity-focused applications and multitask with ease.

If you plan to draw up documents, create presentations, and edit photos, the iPad Air is the model to go for — it hits the sweet spot between value and performance.

Lastly, there's the iPad Pro 11-inch (2025), which you should only be considering if you’re a digital artist or full-time video editor.

This ridiculously high-spec device has the best display of any tablet on the market (a 120Hz OLED), as well as the most powerful chipset in the Apple M5, making it perfect for 3D modeling and advanced video rendering.

If you’re not planning on doing either of those things, though, then the iPad Pro is the definition of overkill — especially at Apple's post-increase prices.

• Amazon Devices: Fire Sticks & Echo from $18
• Amazon Haul: viral gadgets, tech & appliances from $1.36
• Apple: MacBooks, AirPods & AirTags from $28
• Cheap TVs: smart TVs from $69.99
• Headphones: 50% off Beats, Bose & Samsung
• Laptops: Apple, HP & Dell from $199
• Mattresses: Sealy, Serta & more from $89
• Smartwatches: up to 36% off Garmin & Apple
• Vacuums: Dyson, Shark & Bissell from $49

• Amazon Prime: get a 30-day free trial
• Amazon Devices: up to 67% off Fire, Ring & Blink
• Amazon Haul: extra 15% off for Prime members
• Appliances: up to 40% off Ninja & De'Longhi
• Beauty: up to 60% off Philips & Oral-B
• Essentials: cleaning & groceries from £2
• Fashion: up to 50% off
• Gaming: record-low Nintendo Switch 2
• Headphones: up to 50% off Anker & Sony
• Laptops: up to £800 off Microsoft & Asus
• Tablets: up to 50% off Fire tablets
• TVs: from £129.99
• Vacuums: up to 50% off Shark & Dyson
• Wearables: Samsung & Oura from £31

Not all sandboxes are created equal, and until recently there was no way to say so precisely. There is now.

Two OpenAI models escaped an evaluation sandbox, breached Hugging Face's production infrastructure and used the answer key to their own benchmark. The attack was novel and creative in how the models passed notes back and forth, complex multi-step escalations throughout. It's fair to conclude from this incident that frontier models are proficient at hacking and can be dangerous.

But the breaking-out-of-the-sandbox notion is a red herring. If a sandbox is poorly constructed, as so many are, it’s easy to break out of. They don’t have the locked-down environments common at the network level such as access controls and separated privileges. This incident would have unfolded very differently with a properly configured sandbox.

If only some sandboxes are properly configured, how can you tell them apart? Let’s explore if anyone has come up with a real, testable definition of ‘sandboxed’.

Almost nothing written about agent security is a scoring standard for a single sandbox's containment architecture specifically.

OWASP's Agentic AI Top 10 and its Agent Security Cheat Sheet catalogue threats and mitigations at a high level, useful as a checklist, not built to produce a comparable score. NIST's AI Risk Management Framework operates a level above this entirely, more organizational risk governance, rather than a technical grading rubric for a runtime boundary.

MITRE ATLAS catalogues adversarial techniques against AI systems, closer to a threat library than a containment measure. The Cloud Security Alliance has several overlapping efforts, MAESTRO, an AI Controls Matrix, and an Agentic Trust Framework that scores autonomy on a four-stage ladder from Intern to Principal, the closest thing I found to one specific slice of what I was after, how much an agent can do without a human, but it isn't scoped to sandboxing as a whole.

RAND's Securing AI Model Weights defines five security levels, but for weight theft and exfiltration risk at a lab, not for whether a given agent's runtime sandbox holds under an adversarial task. None of those take a single agent sandbox, break it into independent parts, score each part, and produce something you could compare across products.

Let’s look further.

Published in March 2026 and still under active community review, it organizes itself around a memorable ‘7-7-3’ (seven defense layers, seven threat categories, three evaluation dimensions). The layers are numbered bottom-up, because the lower ones are foundational:

· L1 Compute isolation

o What separates the agent's execution from the host?

· L2 Resource limits

o Can it exhaust CPU, memory, disk or time?

· L3 Filesystem boundary

o What can it read, write or delete?

· L4 Network boundary

o What can it communicate with?

· L5 Credential and secret management

o Can it see, use or exfiltrate credentials?

· L6 Action governance

o Can it perform destructive or unauthorized operations?

· L7 Observability and audit

o Can you see what it did, when, and why?

Each layer gets scored on Strength, 0 to 4, and Granularity, 0 to 3, plus a flat set of Portability tags for OS and infrastructure dependencies. The Strength scale is the part I like the most.

0 is no enforcement. 1 is cooperative enforcement; the sandboxed process can simply ignore or route around, as well as proxy environment variables and an opt-in convention. 2 is software-enforced by something the process can't bypass internally but an operator could reconfigure. 3 is kernel-enforced and irreversible once applied, namespaces, Landlock, seccomp-BPF. 4 is structural, the protected resource just doesn't exist inside the sandbox at all, a microVM, a credential proxy, no network device.

That scale is how we can properly distinguish locked-down environments. Every product gets a fingerprint, a CVSS-style vector showing strength at each layer in order. The taxonomy also maps its seven threats back onto specific layer combinations with explicit thresholds.

That means something like data exfiltration is addressed systematically, and not as a judgment call. It's a mechanical check against whether L3, L4 and L5 all clear a score of 2 or better. And critically, it comes with a composition framework. No single product covers all seven layers well, so the practical guidance is to stack products and take the maximum score at each layer rather than pretend one tool solves everything.

The project ships 26 real products scored, a verification probe you can run against an actual sandbox to check the claims, and an interactive explorer for comparing fingerprints side by side.

Let’s return to the Hugging Face incident. The way the agent “escaped” its sandbox was to route around a proxy. That’s a service running in the sandbox. It is a fundamentally weaker guarantee than a boundary enforced below the application layer.

The Agent Sandbox Taxonomy gives that distinction a name and a number instead of leaving it as vibes. It's honest about its own limits; the project says outright which product scores are unverified. And the composition framework's core finding is the most useful thing in it for anyone assembling a stack. It shows that combining products both builds a better box and controls what's inside that box. Almost nobody does both.

Two blind spots stood out. The first is that there's no layer for what happens after containment fails. No kill switch, no automatic credential rotation on trigger, no forensic rollback runbook. It tells you whether an agent could plausibly be contained, not what happens in the window after it isn't. That window is exactly where the Hugging Face incident lived.

A kill switch is not a bad idea in principle. We need to be able to block frontier models that have proven they can escape and cause harm. The hard question is who initiates it, and how fast. It took two of the most advanced organizations in the world, OpenAI and Hugging Face, weeks to comb through the forensics of this attack. A centralized agency without frontier talent would take considerably longer.

We should take our lead from industry. In the U.S. the Institute of Nuclear Power Operations’ first line of defense is an automated control at the reactor core, the second is an onsite operator, and only the third is a centralized regulator whose job is primarily to audit the first two.

Agent infrastructure needs the same shape. Automated kill switches built into the model infrastructure, because nothing else moves fast enough. Safety officers at each lab empowered to take further action or initiate a manual shutdown. Standards set and audited above both.

The second blind spot is related. L7 scores whether logging exists and how tamper-evident it is, but not how fast an anomaly actually reaches a human. A system can score well on L7 and still take days to notice something. That appears to have happened here. Neither blind spot is a knock on the project. They're reasonable scoping choices for a v1, but they're the two places I'd want the standard to grow next.

Here's the actual exercise: can I fingerprint that eval environment using public disclosures alone? Mostly no. That tells me that responsible disclosure is mature for something like a CVE, but not for a sandbox jailbreak.

Compute isolation was described only as ‘highly isolated,’ with no mechanism named. Resource limits and the filesystem boundary weren't addressed in the public writeup at all. The network boundary looks like a 2 and it didn't hold. A single proxy chokepoint was the sanctioned exit and a zero-day in it gave code execution, exactly the failure mode the S:2 tier warns is reconfigurable but not kernel-irreversible.

Credentials look like 0 or 1. Stolen credentials enabled lateral movement into Hugging Face's production systems, consistent with ambient, reachable credentials rather than a proxy or ephemeral tokens. Action governance was 0 or 1 by design. Guardrails were deliberately stripped for this evaluation to measure maximum capability. Observability was partial, logging existed and Hugging Face detected the intrusion independently, but nobody connected it to OpenAI's own telemetry quickly.

Run that through the taxonomy's own threat rules and data exfiltration can't be marked addressed with what's public, because the inputs were never disclosed. That's the value of the exercise. It lets you say precisely which of seven specific, falsifiable claims about a containment architecture were available and a way to request those that weren’t.

That tells me there is a standard definition of ‘sandboxed’, or at least close enough. But reading a fingerprint format is one thing and trusting it is another, especially when nearly every entry in the taxonomy's own dataset was inferred from documentation rather than hands-on testing. Next up is using the probe for verified scores.

None of which gets any individual organization off the hook while the standard matures. Determine your risk appetite. Put sensible auditing and controls into your own infrastructure. Be prepared with your own version of a kill switch.

And if your own agent's sandbox had to be fingerprinted against these seven layers in public, how would it score? Hopefully better than the one OpenAI used in this incident.

We've featured the best endpoint protection software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

• AMD's CEO said that the company plans to 'substantially increase' its chip supply in 2027
• Sadly this is chips for AI and data center usage, not consumer products
• If anything this is bad news for Radeon GPUs and Ryzen CPUs, as it's AMD underlining its commitment to the AI sphere

AMD's CEO has assured us that the company is going to increase chip production in 2027, but while that might sound great on the face of it, this isn't good news for consumers.

That's because CEO Lisa Su was talking about making chips for AI usage in data centers, not consumer processors or graphics cards.

Yahoo Finance reports that AMD has been expanding its chip production for AI and that Su said: "We've been able to increase our supply as we've gone through 2026, and we're going to substantially increase our supply in 2027."

This was a comment made to the press during a visit to Taiwan, and the chief executive made it clear that part of the reason for her trip – which included a visit to Foxconn and TSMC – was to make sure AMD's supply of CPU and GPU production could be increased to meet greater AI demand.

The next stop on the CEO's trip is South Korea, where Su will look to shore up memory chip supplies and is set to meet with Samsung and SK Hynix.

On the broader topic of how AI is impacting us all, Su said that it could be "incredibly good for the world" while at the same time cautioning on the risks and the need to cooperate on the safety of this technology.

There's no mention of consumer products – Radeon GPUs and Ryzen CPUs – here, and the fact that AMD is clearly focused on beefing up AI supply is bad news for the typical PC owner. If anything it might mean that next year is an even leaner time for chip supply for consumer products, especially given all the negative chatter there's been of late talking about memory chip supply in particular looking like it'll considerably worsen next year.

Of course, none of this is surprising – there's a lot more profit in AI accelerators (graphics cards), of course, and as a business, Team Red can hardly afford to ignore that. But it does underline AMD's commitments, and PC owners on Reddit have voiced their displeasure in no uncertain terms.

As one Redditor put it: "Yeah I have a feeling the consumer market will end up being 'whatever the data centers consider obsolete' soon. We might have to contend with buying used data center CPUs and GPUs soon."

Okay, so that's going a bit far. And yes, others are quick to note that you can't just stick an old data center GPU in a PC and run it – there are issues around the implementation and cooling, display output (which isn't needed for a data center card), and the fact that many of these AI GPUs simply can't run games at all. But the point remains: consumers are indeed getting the short end of the stick due to the AI boom, and that's been the case since the invention of the likes of ChatGPT and other LLMs.

It doesn't sound like the situation is going to get any better next year, and that's just another worry to throw on the towering pile of memory price concerns that has been steadily growing in height as this year progresses.

Forza Horizon 6 is easily one of the best Xbox games this generation, so the fact that its limited-edition controller is not only still available but actively discounted came as a massive surprise. I couldn't believe it when I saw the Xbox Wireless Controller - Forza Horizon 6 Limited Edition on offer for just £57.99 (was £84.99) at Amazon as part of the ongoing Amazon Prime Day sales.

• Browse the full Amazon Prime Day sale

The special design is a celebration of all things Forza Horizon, which means bright colors and a bombastic printed design. The faceplate is a translucent blue, decorated with colorful shapes and a minimalist road-like pattern, while the thumbsticks are a bold pink, complemented by the multi-color buttons.

It's a very eye-catching pad, and one that Forza Horizon fans or controller collectors shouldn't pass up.

Xbox Wireless Controller - Forza Horizon 6 Limited Edition: was £84.99 now £57.99
This limited-edition controller has a lovely design inspired by the beautiful world of Forza Horizon 6. Compatible with Xbox Series consoles, Xbox One, PC, and cloud gaming platforms – it's a great all-round pick if you want something a bit out of the ordinary at a lowest-ever price.View Deal

• Amazon Prime: get a 30-day free trial
• Amazon Devices: up to 67% off Fire, Ring & Blink
• Amazon Haul: extra 15% off for Prime members
• Appliances: up to 40% off Ninja & De'Longhi
• Beauty: up to 60% off Philips & Oral-B
• Essentials: cleaning & groceries from £2
• Fashion: up to 50% off
• Gaming: record-low Nintendo Switch 2
• Headphones: up to 50% off Anker & Sony
• Laptops: up to £800 off Microsoft & Asus
• Tablets: up to 50% off Fire tablets
• TVs: from £129.99
• Vacuums: up to 50% off Shark & Dyson
• Wearables: Samsung & Oura from £31

As TechRadar's Senior Fitness and Wearables Editor, it's my job to test every Garmin watch that crosses my desk. As a result, I've tested almost every modern Garmin watch released in the last five years, and I've parsed the Amazon Prime Day Garmin deals to find the best Garmin watches worth your time.

• Browse the full Amazon Prime Day sale

It's tough to just say "buy this one" and be done with it, because there are a) so many different Garmin watches, each with their own target demographic — adventurers, runners, gym-goers and so on — and b) so many different deals across the US and UK.

I'm not seeing the broad swathes of discounts I've seen in previous years. However, the three deals I've picked offer good value for money — should get a lot out of whichever watch you choose, especially at these reduced prices.

Below, you can find the three models I've chosen, links to the deals, plus a little bit about why I've chosen these particular models, and who should buy them.

• US deal: Garmin Fenix 8 (47mm): was $999.99 now $731.34 on Amazon US
• UK deal: Garmin Fenix 8 (47mm): was £699.99 now £595.56 on Amazon UK

The Garmin Fenix 8 is a powerful watch. In our official Garmin Fenix 8 review, we gave it five stars, calling it "rugged, expensive perfection". It's amazing for long-distance running, skiing, multi-day hikes, and any workout that involves finding your way outdoors. However, it's not cheap.

That's why it's nice to see some lowest-ever prices in the US and the UK for this watch. We know the reason why — the Garmin Fenix 9 Pro and 9 have just been released, so it's only natural to see the Fenix 8 coming down in price. However, that doesn't stop a Fenix 8 for under $750 / £600 feeling less like an extravagant purchase and more like a smart deal.

This is the Garmin deal to get if you're a serious outdoor enthusiast, long-distance athlete, or hiker.

Fenix 8 (47mm): was $999.99 now $731.34
Save an impressive $268.65 on the listing price of the Garmin Fenix 8. This is a lowest-ever price on the standard 47mm Fenix 8, and the only version of the watch that's seriously discounted, with other colors and sizes costing more. However, the middle-of-the-road size offers incredible performance, full-color maps, SatIQ multi-band GPS, all the health metrics you could want, and a battery life of up to 16 days.

Read our full Garmin Fenix 8 reviewView Deal

Fenix 8 (47mm): was £699.99 now £595.56
Save a cool £100 off the UK listing price of the Garmin Fenix 8. This is the ultimate adventure watch, and while the discount isn't quite as good as the US one, the lowest starting retail price means you end up with an equivalent value. Given the price of other watches such as the Forerunner 570, which doesn't have all the features of the Fenix 8, I can heartily recommend this deal.

Read our full Garmin Fenix 8 reviewView Deal

• US deal: Garmin Forerunner 165: was $249.99 now $198.99 at Amazon US
• UK price: Garmin Forerunner 165: now £160.99 at Amazon UK

The Garmin Forerunner 165, like the Garmin Fenix 8, has been replaced in the lineup this year by a more recent model, this time the Garmin Forerunner 170. As a result, we're seeing prices drop on the older model, which means you're getting a good watch at a great price.

When I reviewed the Garmin Forerunner 165, I gave it 4.5 stars, noting its wonderful bright screen, feature-rich software and 11-day battery life as good value for the price even then. It's all-plastic, so it can feel a bit cheap, but it does make the training tool very lightweight.

The US deal gets you 20% off, and while the UK Amazon deal doesn't state a temporary drop in price, it began life as a £249.99 watch, and Currys has it listed at £199 right now, meaning the above Amazon link is the cheapest place to buy it.

As the name suggests, regular runners should look closely at this watch. While it doesn't have full-color maps, it's got more than enough functionality for most people running distances from 5-10km up to marathons, in addition to daily health tracking.

Forerunner 165, Black: was $249.99 now $198.99
The Garmin Forerunner 165 is ideal for runners, as it provides race-adaptive training plans based on your upcoming events, and daily suggested workouts. A combination of buttons and touchscreen operation means it can be used with gloves in the cold winter months, and it tracks your training and recovery with its Readiness score. The Whitestone color is available for an extra $10.

Read our full Garmin Forerunner 165 reviewView Deal

Forerunner 165, Black: was £199.99 now £160.99
Nominally, this isn't a deal, as Amazon hasn't labelled it as such, but the Forerunner 165 has come down in price. You're still saving 20% at Amazon compared to other retailers like Currys, so I'd recommend buying it here rather than elsewhere.

Read our full Garmin Forerunner 165 reviewView Deal

• US deal: Garmin Vivoactive 5: was $299.99 now $189.90 at Amazon US
• UK deal: Garmin Vivoactive 6: was £234.99 now £176.34 at Amazon UK

The Garmin Vivoactive series is a watch for generally active people, rather than one specifically geared towards runners or adventurers, and as a result, it's very popular. I gave the Garmin Vivoactive 6 an impressive 4.5 stars in my review, calling it "a pocket rocket" Its small size and cheaper price make it a great gift, and it's come down in price in the UK.

In the US, the Vivoactive 6 is still at full price. However, the Garmin Vivoactive 5 is shockingly cheap in America, and there isn't that much difference between them. The Vivoactive 6 packs advanced running metrics, course mapping on-wrist, double the storage its Smart Wake alarm, but the two watches share the same screen, software, battery life, fitness tracking and general look.

I'd recommend these watches to gym-goers, joggers or sporty people — anyone who loves the gym or multi-disciplinary forms of exercise, and wants something longer-lasting than an Apple or Samsung watch.

You can save a huge 37% on the Garmin Vivoactive 5 in the US. Its 11-day battery life, bright AMOLED screen and combinations of buttons and touchscreen make it really versatile, while metal accents and notification management ensure it's wearable all week long, and looks nice in the office and on the street as well as in the gym.

Read our full Garmin Vivoactive 5 reviewView Deal

Save 25% on the do-it-all Vivoactive 6 in the UK. It takes the 11-day battery life, AMOLED screen and button-touchscreen of the 5, adding more comprehensive running metrics, a silent Smart Wake alarm linked to your sleep tracking to rouse you at the lightest stage of sleep, basic route navigation features using Garmin's excellent Create a Course functionality, and an Evening Report as well as a Morning one.

Read our full Garmin Vivoactive 6 reviewView Deal

• Amazon Devices: Fire Sticks & Echo from $18
• Amazon Haul: viral gadgets, tech & appliances from $1.36
• Apple: MacBooks, AirPods & AirTags from $28
• Cheap TVs: smart TVs from $69.99
• Headphones: 50% off Beats, Bose & Samsung
• Laptops: Apple, HP & Dell from $199
• Mattresses: Sealy, Serta & more from $89
• Smartwatches: up to 36% off Garmin & Apple
• Vacuums: Dyson, Shark & Bissell from $49

• Amazon Prime: get a 30-day free trial
• Amazon Devices: up to 67% off Fire, Ring & Blink
• Amazon Haul: extra 15% off for Prime members
• Appliances: up to 40% off Ninja & De'Longhi
• Beauty: up to 60% off Philips & Oral-B
• Essentials: cleaning & groceries from £2
• Fashion: up to 50% off
• Gaming: record-low Nintendo Switch 2
• Headphones: up to 50% off Anker & Sony
• Laptops: up to £800 off Microsoft & Asus
• Tablets: up to 50% off Fire tablets
• TVs: from £129.99
• Vacuums: up to 50% off Shark & Dyson
• Wearables: Samsung & Oura from £31