×
Asus ROG Zephyrus G14 (2025) Review: Everything To Love For

Asus ROG Zephyrus G14 (2025) Review: Everything To Love For

Tech-news

Gaming laptops are big and bulky. They take up a lot of space in the backpack and are a pain to carry around. While these words are true of almost every gaming laptop, there is one that has consistently defied expectations and managed to pack in performance without the bulk: the Asus ROG Zephyrus G14.

For years, Asus has crammed the most insane components into a chassis no bigger or heavier than an ultrabook, and they have done it again. The new G14 houses the fastest AMD Ryzen 9 HX CPU, along with a choice of high-end 50-series GPUs.

I had the opportunity to test out the G14 on my recent trip to Vietnam, and there isn’t a more perfect test for a portable gaming laptop than traveling. In this review, I’ll share my experience of using the device for over two weeks and carrying it in a backpack to various places.

Asus ROG Zephyrus G14 Review

Hisan Kidwai

Summary

The Asus ROG Zephyrus G14 is the jack of all trades, exceptionally blending performance in a MacBook-like chassis. While it’s a bit more expensive than previous models, the improvements to the screen, the exceptional build quality, and the powerful internals more than make up for any additional costs.

Key Specs & Models

Before I begin the review, I would like to mention the different models. My unit of the G14 came with the top-of-the-line AMD Ryzen 9 HX 370 processor, along with the new RTX 5070 Ti GPU, running at 110W.

In addition, the G14 is available in two more models, each featuring the same processor but different GPUs and storage options. There’s an option to get the RTX 5080 mobile GPU.

Tested Unit Alternate Configurations
Display 14-inch 2880 x 1800 OLED running at 120Hz –
Processor AMD Ryzen AI 9 HX 370 –
GPU Nvidia GeForce RTX 5070 Ti Nvidia GeForce RTX 5080
RAM 32GB LPDDR5X –
Storage 2TB PCIe 4.0 NVMe M.2 SSD 1TB PCIe 4.0 NVMe M.2 SSD
Weight 1.57kg 1.57kg

Design & Hardware

When I first received the G14, I was in the cab to the airport and was scared about how I’d fit it in my tight backpack, which already contained my daily driver, the MacBook Air M1. However, all these fears went away when I took the laptop out of the box and was greeted with a design that is quite possibly the best in its class.

If you’ve seen previous G14s, there isn’t much separating the new one from older models. That said, the design is simply lovely. I recently reviewed the Strix G16, where I described its design as understated, but that was incorrect. This G14 is the definition of understated.

Asus has stuck with the MacBook-like aesthetic, where the laptop is compact and made of super-premium materials. Speaking of in-hand feel, I love the soft-touch metal surface, which makes typing extremely easy and doesn’t attract fingerprints.

Image of the back of the Asus ROG Zephyrus G14

Of course, there is always the temptation to add a few extra inches to improve gaming performance even further, but thankfully, the G14 successfully avoids that. It measures 0.72 inches when closed and weighs approximately 1.6 kg

I could easily carry the G14 across various coffee shops in Ho Chi Minh City, and work without straining my back due to the weight. Also, on the durability front, I accidentally dropped the laptop while it was in my bag, and it sustained no damage whatsoever.

The dot matrix light strip is still there on the back for a bit of flair. Honestly, I still prefer the first time Asus used the matrix display on the back of the G14, which was super cool and allowed you to display various things. Although you can customize this array as well, it’s much more restricted in terms of functionality and drains the battery.

Ports

Image of ports on the left side of the G14
Image of ports on the ride side of the G14

Despite its compact size, Asus hasn’t left any port unturned. That’s because on the left, you get Asus’s charging port, an HDMI 2.1 port, a USB-4 (Type-C) port, one USB 3.2 Gen 2 Type-A port, and a 3.5mm audio jack.

On the other hand, the device features another USB 3.2 Gen 2 Type-C port, which supports power delivery, as well as a USB-A port and a MicroSD card reader.

Keyboard & Trackpad

Image of the keyboard and trackpad of the Asus ROG Zephyrus G14

The Asus G14 features a standard 75% layout keyboard, along with an extra row of macro keys, and I have no complaints about it. The keyboard performs amazingly, with plenty of travel, and the keys are extremely tactile, never feeling mushy. It took me just 30 minutes to get used to the layout, and despite writing this lengthy review on the G14’s keyboard, I experienced no wrist pain whatsoever.

The same praise can also be applied to the trackpad, which is spacious and provides a comfortable navigation experience. The surface is super smooth, and the clicks feel satisfying. That said, if you like clicking on menu items, the force required can be a bit tiring.

Display

Image of the screen on the G14

If I had to use just one word to describe the display on the G14, it would be ‘perfect’. Asus has taken the OLED route and equipped the G14 with a gorgeous 3K OLED 16:10 panel featuring a 120Hz refresh rate. In addition, the panel supports 100% of the DCIP-3 color space and features a peak brightness of 500 nits.

There aren’t many better displays on the market than this one. The media experience is a literal treat for the eyes, as the colors look fantastic, the blacks are inky deep, and the HDR performance is also quite decent. The aspect ratio makes working on the G14 a breeze. And coupled with that 120Hz refresh rate, everything feels super snappy and smooth.

Image of the screen

Yes, competitors like Razer offer full 240Hz support on QHD displays, but it’s important to remember that the jump from 120Hz to 240Hz is only noticeable when playing eSports games.

Performance

Image of the performance on the G14

Time to get into the juicy part of the review. As mentioned, the Asus G14 houses the AMD Ryzen AI 9 HX 370 CPU, which comprises four high-performance Zen 5 cores and eight efficient Zen 5c cores. Besides, my unit came with the Nvidia RTX 5070 Ti GPU, with 12GB of VRAM, and a 110W TDP. You also get 32GB of fast LPDDR5X RAM and 2TB of PCIe 4.0 storage.

As expected, the daily performance of the G14 is just exceptional. With its 120Hz display, the device navigates the UI with zero loading times, even in demanding tasks, and without any hitches. The 32GB RAM is enough to keep everything in memory, without having to reload.

Taking on demanding tasks like video editing and coding, the G14 showed no signs of slowing down and handled editing multiple 4K videos with ease.

Geekbench GPU benchmark result
Geekbench CPU benchmark result

I also ran a series of benchmarks, where the G14 once again shone. In Geekbench 6, the laptop scored a hefty 2841 in the single-core test and 14667 in the multi-core test. In 3D Mark’s Fire Strike test, the G14 managed to reach a respectable 32518 points.

Image of the armoury crate app

Asus has also bundled its Armoury Crate app, which includes a whole host of options. First, you can choose between the different performance modes, depending on the tasks. There’s also an option to customize the LED strip on the back, set different Macro keys, and even make custom performance modes.

Gaming & Thermals

Forza Motorsport benchmark

Every previous G14 model has had to compromise on some gaming performance to achieve its slim form factor. But, with advancements in GPU efficiency and cooling technology, this is no longer the case with the newest G14. The gaming performance is spectacular for almost everyone, and with advancements to DLSS, such as frame generation, the experience is enhanced even further.

Starting with eSports titles like Counter-Strike 2, the G14 knocks it out of the park, delivering over 150 FPS at the maxed-out settings. Moving over to Red Dead Redemption 2, set at 2.5K resolution and High graphics, the G14 maintained a solid 95 FPS average, with 1% lows at 80 FPS.

Cyberpunk 2044 benefits from DLSS the most. That’s because without it, the G14 struggles to achieve over 60 FPS at maximum settings. However, with DLSS frame generation enabled, the frame rates increase to an eye-watering 130 FPS. Of course, some people would argue that DLSS frame generation is not an authentic experience, but at this screen size, you can’t really tell the difference. The smoothness of playing at over 100 FPS offsets any visual anomalies.

Asus ROG zephyrus G14 Indiana Jones game settings
Indiana Jones Graphics Settings

Moving over to Forza Motorsport, I cranked everything up to maximum and turned RTX reflections on for everything. The result was an average of 49 FPS. Finally, in Indiana Jones and The Great Circle at Ultra graphics settings and DLSS enabled, I achieved an average of over 100 FPS, with 1% lows hovering around 90 FPS.

Asus claims it has significantly improved the thermal performance of previous generations, and it shows. I experienced no thermal throttling during my gaming sessions, and all temperatures remained in check. However, it’s worth noting that the underside of the laptop can become quite toasty when gaming, but that’s expected from every gaming laptop.

Battery Life & Charging

g14 battery stats

Battery life is often the Achilles heel when it comes to laptops, and despite the efficiency improvements, the same is true in the case of the G14. Don’t get me wrong, the G14 performs better than its predecessors, but still, the 4-5 hours of battery life on Silent mode isn’t that great. I primarily worked on Chrome, did some research, and watched a few videos. When gaming on battery, the battery life lasts around an hour.

That said, the bulky charger does an excellent job of fully topping up the battery in just over an hour, from 20%.

Webcam & Speakers

Image of the webcam performance

Like the Strix, the 1080p webcam on the G14 is pretty decent. Give it enough light, and the video quality is pretty good with natural colors and some HDR, too. As expected, the quality falls short in low lighting, but that’s not much to worry about.

On the other hand, however, I was hugely impressed with the speakers on the G14. That’s because instead of putting them underneath the chassis, Asus has made the wise decision of putting them on top, and that changes the whole game. Gone is the muddiness when keeping the laptop on a soft surface, and it’s replaced by crisp sound, with clear vocals, amazing bass, and treble.

Should You Buy The Asus ROG Zephyrus G14?

Asus ROG Zephyrus G14 design image

If you could only buy one laptop for both work and gaming, I say look no further than the Asus ROG Zephyrus G14. When it was released years ago, there was nothing like it, and even now, in 2025, there is still no laptop like this. Asus has managed to blend performance and form factor like no other laptop maker, and its latest refresh is probably the best one.

Sure, at INR 1,84,999 or USD 2,499, it’s a bit more expensive this time, but with the improvements to the screen, the exceptional build quality, and the powerful internals, I feel it’s worth it. It’s an investment, and if you’re in the market for such a laptop, the Zephyrus G14 is the jack of all trades, and you should definitely check it out.

Source link
#Asus #ROG #Zephyrus #G14 #Review #Love

Tag

Related Posts

The world could face the strongest El Niño since recordkeeping began, increasing the likelihood of extreme weather events and significant, long-term economic impacts globally.

A new analysis by Berkeley Earth climate scientist Zeke Hausfather paints a worrying picture of the climate phenomenon developing in the Pacific Ocean. Models have consistently forecast that this year’s El Niño—characterized by hotter than normal waters in the eastern tropical Pacific—could be among the strongest on records. Hausfather’s analysis using the El Niño forecast based on 14 climate models through July indicates that “this year’s El Niño is not only very likely to be the strongest event since reliable records began—it may end up the strongest by a truly mind-blowing margin.”

Climate researchers measure the strength of El Niño by looking at the temperatures in a region of the Pacific known as Niño 3.4, which covers the eastern and central part of the ocean basin. The new analysis relies on 667 simulations and compares them with historical records of the most intense El Niño events since 1877. It also disentangles El Niño’s intensity from the background warming driven by fossil fuels, making it easier to draw comparisons between super El Niños.

The model average shows that heat in the Niño 3.4 region could peak at 3.6 degrees Celsius (6.5 degrees Fahrenheit) above normal levels. This figure would exceed the previous record—set during the 2015-2016 El Niño episode—by approximately 0.8 degrees Celsius. It’s rare for records to fall by such a wide spread.

“For context, the gap between the strongest and the fifth strongest El Niño of the past 150 years is only about 0.5C. The models are forecasting something outside the envelope of anything we have ever observed,” Hausfather writes.

Of course, what happens in the tropical Pacific doesn’t stay in the tropical Pacific. El Niño rearranges the atmosphere, changing weather patterns around the world. The more intense the El Niño, the more intense the impacts (usually).

Record-breaking ocean heat in the Pacific combined with warming driven by carbon emissions could send the global temperature to new highs. But it takes a while for the extra heat released by El Niño to work its way through the vast climate system, meaning that next year is when we’ll see the full impact of this year’s Pacific heat wave on the global average temperature. According to his projections, global temperatures in 2027 could reach up to 1.7 degrees Celsius above the preindustrial average, well above the 1.5 degrees Celsius threshold that’s been dubbed a relatively safe level of global warming.

But the world won’t have to wait until 2027 to feel the heat of El Niño. A separate analysis by Carbon Brief shows that the odds are increasing for 2026 to set an annual heat record, with the likelihood up from 19 percent in April to 35 percent as of July. And the forecasters are expecting extreme weather associated with El Niño to impact various parts of the globe in the coming months. In fact, some impacts are already here: Fisheries in the eastern Pacific are seeing lower catches, with the Peruvian government setting up a ban for catching anchovies this spring.

Hausfather’s analysis also revealed that the phenomenon is developing faster than usual. According to the comparison, this year’s El Niño is intensifying more rapidly than the 1997-1998 event, another super El Niño. It also differs significantly from the 2015 event, which began when the ocean was already showing signs of prior warming—in other words, it had a running start. That’s particularly striking because this year began with La Niña, the cooler-than-normal cousin of El Niño, so it had a bit of work to do before it even got going.

#Niño #Track #Strongest #Recordclimate,environment,weather,el nino">The 2026 El Niño Is on Track to Be the Strongest on RecordThe world could face the strongest El Niño since recordkeeping began, increasing the likelihood of extreme weather events and significant, long-term economic impacts globally.A new analysis by Berkeley Earth climate scientist Zeke Hausfather paints a worrying picture of the climate phenomenon developing in the Pacific Ocean. Models have consistently forecast that this year’s El Niño—characterized by hotter than normal waters in the eastern tropical Pacific—could be among the strongest on records. Hausfather’s analysis using the El Niño forecast based on 14 climate models through July indicates that “this year’s El Niño is not only very likely to be the strongest event since reliable records began—it may end up the strongest by a truly mind-blowing margin.”Climate researchers measure the strength of El Niño by looking at the temperatures in a region of the Pacific known as Niño 3.4, which covers the eastern and central part of the ocean basin. The new analysis relies on 667 simulations and compares them with historical records of the most intense El Niño events since 1877. It also disentangles El Niño’s intensity from the background warming driven by fossil fuels, making it easier to draw comparisons between super El Niños.The model average shows that heat in the Niño 3.4 region could peak at 3.6 degrees Celsius (6.5 degrees Fahrenheit) above normal levels. This figure would exceed the previous record—set during the 2015-2016 El Niño episode—by approximately 0.8 degrees Celsius. It’s rare for records to fall by such a wide spread.“For context, the gap between the strongest and the fifth strongest El Niño of the past 150 years is only about 0.5C. The models are forecasting something outside the envelope of anything we have ever observed,” Hausfather writes.Of course, what happens in the tropical Pacific doesn’t stay in the tropical Pacific. El Niño rearranges the atmosphere, changing weather patterns around the world. The more intense the El Niño, the more intense the impacts (usually).Record-breaking ocean heat in the Pacific combined with warming driven by carbon emissions could send the global temperature to new highs. But it takes a while for the extra heat released by El Niño to work its way through the vast climate system, meaning that next year is when we’ll see the full impact of this year’s Pacific heat wave on the global average temperature. According to his projections, global temperatures in 2027 could reach up to 1.7 degrees Celsius above the preindustrial average, well above the 1.5 degrees Celsius threshold that’s been dubbed a relatively safe level of global warming.But the world won’t have to wait until 2027 to feel the heat of El Niño. A separate analysis by Carbon Brief shows that the odds are increasing for 2026 to set an annual heat record, with the likelihood up from 19 percent in April to 35 percent as of July. And the forecasters are expecting extreme weather associated with El Niño to impact various parts of the globe in the coming months. In fact, some impacts are already here: Fisheries in the eastern Pacific are seeing lower catches, with the Peruvian government setting up a ban for catching anchovies this spring.Hausfather’s analysis also revealed that the phenomenon is developing faster than usual. According to the comparison, this year’s El Niño is intensifying more rapidly than the 1997-1998 event, another super El Niño. It also differs significantly from the 2015 event, which began when the ocean was already showing signs of prior warming—in other words, it had a running start. That’s particularly striking because this year began with La Niña, the cooler-than-normal cousin of El Niño, so it had a bit of work to do before it even got going.#Niño #Track #Strongest #Recordclimate,environment,weather,el nino

strongest El Niño since recordkeeping began, increasing the likelihood of extreme weather events and significant, long-term economic impacts globally.

A new analysis by Berkeley Earth climate scientist Zeke Hausfather paints a worrying picture of the climate phenomenon developing in the Pacific Ocean. Models have consistently forecast that this year’s El Niño—characterized by hotter than normal waters in the eastern tropical Pacific—could be among the strongest on records. Hausfather’s analysis using the El Niño forecast based on 14 climate models through July indicates that “this year’s El Niño is not only very likely to be the strongest event since reliable records began—it may end up the strongest by a truly mind-blowing margin.”

Climate researchers measure the strength of El Niño by looking at the temperatures in a region of the Pacific known as Niño 3.4, which covers the eastern and central part of the ocean basin. The new analysis relies on 667 simulations and compares them with historical records of the most intense El Niño events since 1877. It also disentangles El Niño’s intensity from the background warming driven by fossil fuels, making it easier to draw comparisons between super El Niños.

The model average shows that heat in the Niño 3.4 region could peak at 3.6 degrees Celsius (6.5 degrees Fahrenheit) above normal levels. This figure would exceed the previous record—set during the 2015-2016 El Niño episode—by approximately 0.8 degrees Celsius. It’s rare for records to fall by such a wide spread.

“For context, the gap between the strongest and the fifth strongest El Niño of the past 150 years is only about 0.5C. The models are forecasting something outside the envelope of anything we have ever observed,” Hausfather writes.

Of course, what happens in the tropical Pacific doesn’t stay in the tropical Pacific. El Niño rearranges the atmosphere, changing weather patterns around the world. The more intense the El Niño, the more intense the impacts (usually).

Record-breaking ocean heat in the Pacific combined with warming driven by carbon emissions could send the global temperature to new highs. But it takes a while for the extra heat released by El Niño to work its way through the vast climate system, meaning that next year is when we’ll see the full impact of this year’s Pacific heat wave on the global average temperature. According to his projections, global temperatures in 2027 could reach up to 1.7 degrees Celsius above the preindustrial average, well above the 1.5 degrees Celsius threshold that’s been dubbed a relatively safe level of global warming.

But the world won’t have to wait until 2027 to feel the heat of El Niño. A separate analysis by Carbon Brief shows that the odds are increasing for 2026 to set an annual heat record, with the likelihood up from 19 percent in April to 35 percent as of July. And the forecasters are expecting extreme weather associated with El Niño to impact various parts of the globe in the coming months. In fact, some impacts are already here: Fisheries in the eastern Pacific are seeing lower catches, with the Peruvian government setting up a ban for catching anchovies this spring.

Hausfather’s analysis also revealed that the phenomenon is developing faster than usual. According to the comparison, this year’s El Niño is intensifying more rapidly than the 1997-1998 event, another super El Niño. It also differs significantly from the 2015 event, which began when the ocean was already showing signs of prior warming—in other words, it had a running start. That’s particularly striking because this year began with La Niña, the cooler-than-normal cousin of El Niño, so it had a bit of work to do before it even got going.

#Niño #Track #Strongest #Recordclimate,environment,weather,el nino">The 2026 El Niño Is on Track to Be the Strongest on Record

The world could face the strongest El Niño since recordkeeping began, increasing the likelihood of extreme weather events and significant, long-term economic impacts globally.

A new analysis by Berkeley Earth climate scientist Zeke Hausfather paints a worrying picture of the climate phenomenon developing in the Pacific Ocean. Models have consistently forecast that this year’s El Niño—characterized by hotter than normal waters in the eastern tropical Pacific—could be among the strongest on records. Hausfather’s analysis using the El Niño forecast based on 14 climate models through July indicates that “this year’s El Niño is not only very likely to be the strongest event since reliable records began—it may end up the strongest by a truly mind-blowing margin.”

Climate researchers measure the strength of El Niño by looking at the temperatures in a region of the Pacific known as Niño 3.4, which covers the eastern and central part of the ocean basin. The new analysis relies on 667 simulations and compares them with historical records of the most intense El Niño events since 1877. It also disentangles El Niño’s intensity from the background warming driven by fossil fuels, making it easier to draw comparisons between super El Niños.

The model average shows that heat in the Niño 3.4 region could peak at 3.6 degrees Celsius (6.5 degrees Fahrenheit) above normal levels. This figure would exceed the previous record—set during the 2015-2016 El Niño episode—by approximately 0.8 degrees Celsius. It’s rare for records to fall by such a wide spread.

“For context, the gap between the strongest and the fifth strongest El Niño of the past 150 years is only about 0.5C. The models are forecasting something outside the envelope of anything we have ever observed,” Hausfather writes.

Of course, what happens in the tropical Pacific doesn’t stay in the tropical Pacific. El Niño rearranges the atmosphere, changing weather patterns around the world. The more intense the El Niño, the more intense the impacts (usually).

Record-breaking ocean heat in the Pacific combined with warming driven by carbon emissions could send the global temperature to new highs. But it takes a while for the extra heat released by El Niño to work its way through the vast climate system, meaning that next year is when we’ll see the full impact of this year’s Pacific heat wave on the global average temperature. According to his projections, global temperatures in 2027 could reach up to 1.7 degrees Celsius above the preindustrial average, well above the 1.5 degrees Celsius threshold that’s been dubbed a relatively safe level of global warming.

But the world won’t have to wait until 2027 to feel the heat of El Niño. A separate analysis by Carbon Brief shows that the odds are increasing for 2026 to set an annual heat record, with the likelihood up from 19 percent in April to 35 percent as of July. And the forecasters are expecting extreme weather associated with El Niño to impact various parts of the globe in the coming months. In fact, some impacts are already here: Fisheries in the eastern Pacific are seeing lower catches, with the Peruvian government setting up a ban for catching anchovies this spring.

Hausfather’s analysis also revealed that the phenomenon is developing faster than usual. According to the comparison, this year’s El Niño is intensifying more rapidly than the 1997-1998 event, another super El Niño. It also differs significantly from the 2015 event, which began when the ocean was already showing signs of prior warming—in other words, it had a running start. That’s particularly striking because this year began with La Niña, the cooler-than-normal cousin of El Niño, so it had a bit of work to do before it even got going.

#Niño #Track #Strongest #Recordclimate,environment,weather,el nino

For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.

Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.

The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings. 

Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today.

Meta launches Seller, a standalone app for Facebook Marketplace sellers For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings.  Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today. Mashable Light Speed Credit: Facebook Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators. As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform.  #Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers

Credit: Facebook

Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. 

Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.

In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators.

As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.

Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform. 

#Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers">Meta launches Seller, a standalone app for Facebook Marketplace sellers For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings.  Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today. Mashable Light Speed Credit: Facebook Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators. As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform.  #Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers

releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.

The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings. 

Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today.

Meta launches Seller, a standalone app for Facebook Marketplace sellers For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings.  Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today. Mashable Light Speed Credit: Facebook Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators. As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform.  #Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers

Credit: Facebook

Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. 

Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.

In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators.

As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.

Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform. 

#Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers">Meta launches Seller, a standalone app for Facebook Marketplace sellers

For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.

Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.

The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings. 

Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today.

Meta launches Seller, a standalone app for Facebook Marketplace sellers For some social media users, Facebook may feel like an app left behind in an online era of the past. However, for others, Facebook has reinvented itself as the modern-day Craigslist with Facebook Marketplace.Meta seems to have taken notice and is now releasing a standalone app for the Facebook Marketplace power users who keep the platform going: It’s sellers.The new app, aptly named Seller, launched on Friday and provides Facebook Marketplace sellers with tools to list their wares and manage their current listings.  Like Facebook Marketplace itself, the Seller app is free for users. Meta monetizes Facebook Marketplace with ads and does not take any percentage of its users’ sales. Users will be able to download the app and login with their Facebook account starting today. Mashable Light Speed Credit: Facebook Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators. As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform.  #Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers

Credit: Facebook

Seller provides a tool to post new listings, a dashboard to manage their existing ones, an in-app inbox to reply to prospective buyers, and analytics for all of their old listings. 

Seller also introduces a few new tools for sellers as well. According to Facebook, Seller users will be able to scan their products with their smartphone, and AI will fill-out the listing details automatically.

In an interview with the New York Times, Facebook head Tom Allison said that the app would be the Marketplace seller equivalent of a video editing tool for content creators. Allison also explained how Meta is experimenting with more standalone products for Facebook’s power users, which include Marketplace Sellers and Facebook Groups administrators.

As social media users move to competing platforms like TikTok and Meta’s own Instagram, Facebook has found new life within specific, once-niche features on the platform. In May, Facebook launched a standalone app called Forum for its Facebook Groups power users. Now, the company is doing the same for Facebook Marketplace sellers.

Meta says that one-third of young adults who use Facebook also use Marketplace. And, every month, more than 420 million items are listed on the Marketplace platform. 

#Meta #launches #Seller #standalone #app #Facebook #Marketplace #sellers
For months, AI giants have devised special vetted programs and strict guardrails to limit the use of their models by malicious hackers. But these limits are now hindering the work of legitimate network defenders, as well as that of offensive cybersecurity researchers. 

In June, the U.S. government slapped export control restrictions on Anthropic’s much-hyped AI models Mythos and Fable. The move was prompted at least in part by a report that claimed it was possible to bypass the models’ guardrails designed to prevent users from using them to build and execute malicious cyberattacks.

Regardless of whether the incident was really motivated by fears of a jailbreak, the fact is that Anthropic has repeatedly marketed Mythos as some kind of doomsday cybermachine that can only be given to carefully vetted users, and even then with strict guardrails in place. (The export controls on Fable 5 and Mythos 5 have since been lifted. Fable 5 returned to general access on July 1; Mythos 5 has been reintroduced only to vetted U.S. organizations as part of the government’s review process.)

That kind of gatekeeping isn’t unique to Mythos. Both Anthropic, with its other models, and OpenAI offer cybersecurity researchers programs they can apply to get vetted and — if approved — access models with fewer cybersecurity restrictions: OpenAI’s Trusted Access for Cyber program and Anthropic’s Cyber Verification Program. 

These guardrails have been widely criticized, particularly by researchers whose job is to find unknown vulnerabilities in systems and devise ways to exploit them before criminals do.

During a recent appearance on a cybersecurity podcast, Mark Dowd, a well-known security researcher, said that, “it’s not really comfortable to me that these random large companies are making arbitrary decisions about what is safe in security and what’s not.”

Dowd has spent decades finding and selling “zero-days” — previously unknown software flaws and the exploits that take advantage of them — to Western governments, rather than reporting them to the software makers so they get patched. Governments pay a premium for vulnerabilities precisely because they stay open, which is useful for intelligence operations.

Dowd admitted his work may make him biased, but he isn’t alone. Several people who work in offensive cybersecurity — they proactively probe systems for weaknesses — described to TechCrunch how they use AI tools and deal with their guardrails. 

Chris Anley, the chief scientist at security consulting giant NCC Group, said that asking an AI model to try to exploit a bug is a key step in confirming it’s a real vulnerability worth fixing. But if a guardrail prompts the model to refuse to answer the question outright, the guardrail hurts defenders, he said.

“This is where the whole offensive versus defensive and guardrails part comes in, because ‘fix this code’ as a prompt is both an essential mechanism for defense but also a roadmap for finding critical vulnerabilities in the code base,” said Anley. “So at the same time, the same tool is both an offensive tool and a defensive tool, and the two can’t really be unpicked.”

It’s “like a hammer,” he continued. “You can’t build a house without a hammer. It’s definitely a tool but it’s also irreducibly a weapon as well.”

When he and his colleagues run into such a roadblock, they sometimes fall back on open source AI models that come with no guardrails at all.

Paolo Stagno, the chief technology officer at Crowdfense, a well-known company that develops, acquires, and sells unknown vulnerabilities to government agencies, agreed with Dowd, saying AI companies “essentially treat customers like children who need babysitting” with their vetted programs and guardrails. 

Stagno said he and his colleagues do use frontier models — but only for reverse engineering. They avoid using AI to help find vulnerabilities or build exploits, he said, because feeding that work into a cloud-based model risks leaking sensitive vulnerability data or having it absorbed into future training runs. For that step, he said, they use open source models run locally, as they do not rely on sharing data outside of the model. 

Giuseppe Cali, a security researcher who finds zero-days and develops exploits, said guardrails are not impeding his work. That’s because he doesn’t use AI for offensive work; instead, he uses it for initial reverse engineering, to understand the code he’s analyzing, and to build supporting tools. For that, he said, AI tools can speed up the process and allow him to focus on discovering vulnerabilities. 

“I still want to own the actual bug discovery and weaponization myself and that wouldn’t change if all guardrails were lifted tomorrow,” said Cali. “I am jealous of my bugs, and I like this game too much to let models play it for me.”

One researcher at a smartphone-component manufacturer, who spoke on condition of anonymity because he isn’t authorized to talk to the press, said his employer isn’t part of Anthropic’s CVP program and as a result, its tools are barely useful for finding vulnerabilities because the guardrails are too strict.

“If it catches wind we’re doing anything security related, it just stops and isn’t usable,” the person said. 

Chris Thompson — chief executive of cybersecurity firm RemoteThreat and founder of Offensive AI Con, an offensive security and AI-focused event — said that in his experience using the frontier AI models, the guardrails can be inconsistent and work differently every day. That’s true even inside the looser boundaries of Anthropic’s and OpenAI’s vetted programs. 

“I think the practical impact is you spend a lot of time negotiating with the model instead of working on the core security program,” said Thompson. “Instead of analyzing a vulnerability and reasoning through the exploitability, you’re trying to find why you’re getting inconsistent results or why are models over-sanitizing the output.” 

Consequently, researchers rely on or get pushed toward Chinese open source models like GLM — freely downloadable models that can be run locally with no vetting or usage restrictions — said Thompson.

“You have these responsible researchers that are being pushed away from U.S.-governed systems to foreign-owned systems,” he said. “I think it’s more harmful than good to have these guardrails in place.”

Rather than tightening restrictions further, Thompson called for the AI frontier labs to open up their programs, provide responsible access, and hold those who abuse their tools accountable. Otherwise, he argued, defenders will lose the AI race.

“There’s this big storm coming. There’s this big wave of attacks that are going to happen at speed and scale like never before,” said Thompson. “But the same security consulting firms and legit researchers that are trying to make a difference are being stifled right now.”

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#guardrails #impeding #work #offensive #cybersecurity #researchers #TechCrunchcybersecurity,Zero-days">How AI guardrails are impeding the work of offensive cybersecurity researchers | TechCrunch For months, AI giants have devised special vetted programs and strict guardrails to limit the use of their models by malicious hackers. But these limits are now hindering the work of legitimate network defenders, as well as that of offensive cybersecurity researchers.  In June, the U.S. government slapped export control restrictions on Anthropic’s much-hyped AI models Mythos and Fable. The move was prompted at least in part by a report that claimed it was possible to bypass the models’ guardrails designed to prevent users from using them to build and execute malicious cyberattacks. Regardless of whether the incident was really motivated by fears of a jailbreak, the fact is that Anthropic has repeatedly marketed Mythos as some kind of doomsday cybermachine that can only be given to carefully vetted users, and even then with strict guardrails in place. (The export controls on Fable 5 and Mythos 5 have since been lifted. Fable 5 returned to general access on July 1; Mythos 5 has been reintroduced only to vetted U.S. organizations as part of the government’s review process.) That kind of gatekeeping isn’t unique to Mythos. Both Anthropic, with its other models, and OpenAI offer cybersecurity researchers programs they can apply to get vetted and — if approved — access models with fewer cybersecurity restrictions: OpenAI’s Trusted Access for Cyber program and Anthropic’s Cyber Verification Program.  These guardrails have been widely criticized, particularly by researchers whose job is to find unknown vulnerabilities in systems and devise ways to exploit them before criminals do. During a recent appearance on a cybersecurity podcast, Mark Dowd, a well-known security researcher, said that, “it’s not really comfortable to me that these random large companies are making arbitrary decisions about what is safe in security and what’s not.” Dowd has spent decades finding and selling “zero-days” — previously unknown software flaws and the exploits that take advantage of them — to Western governments, rather than reporting them to the software makers so they get patched. Governments pay a premium for vulnerabilities precisely because they stay open, which is useful for intelligence operations. Dowd admitted his work may make him biased, but he isn’t alone. Several people who work in offensive cybersecurity — they proactively probe systems for weaknesses — described to TechCrunch how they use AI tools and deal with their guardrails.  Chris Anley, the chief scientist at security consulting giant NCC Group, said that asking an AI model to try to exploit a bug is a key step in confirming it’s a real vulnerability worth fixing. But if a guardrail prompts the model to refuse to answer the question outright, the guardrail hurts defenders, he said. “This is where the whole offensive versus defensive and guardrails part comes in, because ‘fix this code’ as a prompt is both an essential mechanism for defense but also a roadmap for finding critical vulnerabilities in the code base,” said Anley. “So at the same time, the same tool is both an offensive tool and a defensive tool, and the two can’t really be unpicked.” It’s “like a hammer,” he continued. “You can’t build a house without a hammer. It’s definitely a tool but it’s also irreducibly a weapon as well.” When he and his colleagues run into such a roadblock, they sometimes fall back on open source AI models that come with no guardrails at all. Paolo Stagno, the chief technology officer at Crowdfense, a well-known company that develops, acquires, and sells unknown vulnerabilities to government agencies, agreed with Dowd, saying AI companies “essentially treat customers like children who need babysitting” with their vetted programs and guardrails.  Stagno said he and his colleagues do use frontier models — but only for reverse engineering. They avoid using AI to help find vulnerabilities or build exploits, he said, because feeding that work into a cloud-based model risks leaking sensitive vulnerability data or having it absorbed into future training runs. For that step, he said, they use open source models run locally, as they do not rely on sharing data outside of the model.  Giuseppe Cali, a security researcher who finds zero-days and develops exploits, said guardrails are not impeding his work. That’s because he doesn’t use AI for offensive work; instead, he uses it for initial reverse engineering, to understand the code he’s analyzing, and to build supporting tools. For that, he said, AI tools can speed up the process and allow him to focus on discovering vulnerabilities.  “I still want to own the actual bug discovery and weaponization myself and that wouldn’t change if all guardrails were lifted tomorrow,” said Cali. “I am jealous of my bugs, and I like this game too much to let models play it for me.” One researcher at a smartphone-component manufacturer, who spoke on condition of anonymity because he isn’t authorized to talk to the press, said his employer isn’t part of Anthropic’s CVP program and as a result, its tools are barely useful for finding vulnerabilities because the guardrails are too strict. “If it catches wind we’re doing anything security related, it just stops and isn’t usable,” the person said.  Chris Thompson — chief executive of cybersecurity firm RemoteThreat and founder of Offensive AI Con, an offensive security and AI-focused event — said that in his experience using the frontier AI models, the guardrails can be inconsistent and work differently every day. That’s true even inside the looser boundaries of Anthropic’s and OpenAI’s vetted programs.  “I think the practical impact is you spend a lot of time negotiating with the model instead of working on the core security program,” said Thompson. “Instead of analyzing a vulnerability and reasoning through the exploitability, you’re trying to find why you’re getting inconsistent results or why are models over-sanitizing the output.”  Consequently, researchers rely on or get pushed toward Chinese open source models like GLM — freely downloadable models that can be run locally with no vetting or usage restrictions — said Thompson. “You have these responsible researchers that are being pushed away from U.S.-governed systems to foreign-owned systems,” he said. “I think it’s more harmful than good to have these guardrails in place.” Rather than tightening restrictions further, Thompson called for the AI frontier labs to open up their programs, provide responsible access, and hold those who abuse their tools accountable. Otherwise, he argued, defenders will lose the AI race. “There’s this big storm coming. There’s this big wave of attacks that are going to happen at speed and scale like never before,” said Thompson. “But the same security consulting firms and legit researchers that are trying to make a difference are being stifled right now.” When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.#guardrails #impeding #work #offensive #cybersecurity #researchers #TechCrunchcybersecurity,Zero-days

slapped export control restrictions on Anthropic’s much-hyped AI models Mythos and Fable. The move was prompted at least in part by a report that claimed it was possible to bypass the models’ guardrails designed to prevent users from using them to build and execute malicious cyberattacks.

Regardless of whether the incident was really motivated by fears of a jailbreak, the fact is that Anthropic has repeatedly marketed Mythos as some kind of doomsday cybermachine that can only be given to carefully vetted users, and even then with strict guardrails in place. (The export controls on Fable 5 and Mythos 5 have since been lifted. Fable 5 returned to general access on July 1; Mythos 5 has been reintroduced only to vetted U.S. organizations as part of the government’s review process.)

That kind of gatekeeping isn’t unique to Mythos. Both Anthropic, with its other models, and OpenAI offer cybersecurity researchers programs they can apply to get vetted and — if approved — access models with fewer cybersecurity restrictions: OpenAI’s Trusted Access for Cyber program and Anthropic’s Cyber Verification Program. 

These guardrails have been widely criticized, particularly by researchers whose job is to find unknown vulnerabilities in systems and devise ways to exploit them before criminals do.

During a recent appearance on a cybersecurity podcast, Mark Dowd, a well-known security researcher, said that, “it’s not really comfortable to me that these random large companies are making arbitrary decisions about what is safe in security and what’s not.”

Dowd has spent decades finding and selling “zero-days” — previously unknown software flaws and the exploits that take advantage of them — to Western governments, rather than reporting them to the software makers so they get patched. Governments pay a premium for vulnerabilities precisely because they stay open, which is useful for intelligence operations.

Dowd admitted his work may make him biased, but he isn’t alone. Several people who work in offensive cybersecurity — they proactively probe systems for weaknesses — described to TechCrunch how they use AI tools and deal with their guardrails. 

Chris Anley, the chief scientist at security consulting giant NCC Group, said that asking an AI model to try to exploit a bug is a key step in confirming it’s a real vulnerability worth fixing. But if a guardrail prompts the model to refuse to answer the question outright, the guardrail hurts defenders, he said.

“This is where the whole offensive versus defensive and guardrails part comes in, because ‘fix this code’ as a prompt is both an essential mechanism for defense but also a roadmap for finding critical vulnerabilities in the code base,” said Anley. “So at the same time, the same tool is both an offensive tool and a defensive tool, and the two can’t really be unpicked.”

It’s “like a hammer,” he continued. “You can’t build a house without a hammer. It’s definitely a tool but it’s also irreducibly a weapon as well.”

When he and his colleagues run into such a roadblock, they sometimes fall back on open source AI models that come with no guardrails at all.

Paolo Stagno, the chief technology officer at Crowdfense, a well-known company that develops, acquires, and sells unknown vulnerabilities to government agencies, agreed with Dowd, saying AI companies “essentially treat customers like children who need babysitting” with their vetted programs and guardrails. 

Stagno said he and his colleagues do use frontier models — but only for reverse engineering. They avoid using AI to help find vulnerabilities or build exploits, he said, because feeding that work into a cloud-based model risks leaking sensitive vulnerability data or having it absorbed into future training runs. For that step, he said, they use open source models run locally, as they do not rely on sharing data outside of the model. 

Giuseppe Cali, a security researcher who finds zero-days and develops exploits, said guardrails are not impeding his work. That’s because he doesn’t use AI for offensive work; instead, he uses it for initial reverse engineering, to understand the code he’s analyzing, and to build supporting tools. For that, he said, AI tools can speed up the process and allow him to focus on discovering vulnerabilities. 

“I still want to own the actual bug discovery and weaponization myself and that wouldn’t change if all guardrails were lifted tomorrow,” said Cali. “I am jealous of my bugs, and I like this game too much to let models play it for me.”

One researcher at a smartphone-component manufacturer, who spoke on condition of anonymity because he isn’t authorized to talk to the press, said his employer isn’t part of Anthropic’s CVP program and as a result, its tools are barely useful for finding vulnerabilities because the guardrails are too strict.

“If it catches wind we’re doing anything security related, it just stops and isn’t usable,” the person said. 

Chris Thompson — chief executive of cybersecurity firm RemoteThreat and founder of Offensive AI Con, an offensive security and AI-focused event — said that in his experience using the frontier AI models, the guardrails can be inconsistent and work differently every day. That’s true even inside the looser boundaries of Anthropic’s and OpenAI’s vetted programs. 

“I think the practical impact is you spend a lot of time negotiating with the model instead of working on the core security program,” said Thompson. “Instead of analyzing a vulnerability and reasoning through the exploitability, you’re trying to find why you’re getting inconsistent results or why are models over-sanitizing the output.” 

Consequently, researchers rely on or get pushed toward Chinese open source models like GLM — freely downloadable models that can be run locally with no vetting or usage restrictions — said Thompson.

“You have these responsible researchers that are being pushed away from U.S.-governed systems to foreign-owned systems,” he said. “I think it’s more harmful than good to have these guardrails in place.”

Rather than tightening restrictions further, Thompson called for the AI frontier labs to open up their programs, provide responsible access, and hold those who abuse their tools accountable. Otherwise, he argued, defenders will lose the AI race.

“There’s this big storm coming. There’s this big wave of attacks that are going to happen at speed and scale like never before,” said Thompson. “But the same security consulting firms and legit researchers that are trying to make a difference are being stifled right now.”

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#guardrails #impeding #work #offensive #cybersecurity #researchers #TechCrunchcybersecurity,Zero-days">How AI guardrails are impeding the work of offensive cybersecurity researchers | TechCrunch

For months, AI giants have devised special vetted programs and strict guardrails to limit the use of their models by malicious hackers. But these limits are now hindering the work of legitimate network defenders, as well as that of offensive cybersecurity researchers. 

In June, the U.S. government slapped export control restrictions on Anthropic’s much-hyped AI models Mythos and Fable. The move was prompted at least in part by a report that claimed it was possible to bypass the models’ guardrails designed to prevent users from using them to build and execute malicious cyberattacks.

Regardless of whether the incident was really motivated by fears of a jailbreak, the fact is that Anthropic has repeatedly marketed Mythos as some kind of doomsday cybermachine that can only be given to carefully vetted users, and even then with strict guardrails in place. (The export controls on Fable 5 and Mythos 5 have since been lifted. Fable 5 returned to general access on July 1; Mythos 5 has been reintroduced only to vetted U.S. organizations as part of the government’s review process.)

That kind of gatekeeping isn’t unique to Mythos. Both Anthropic, with its other models, and OpenAI offer cybersecurity researchers programs they can apply to get vetted and — if approved — access models with fewer cybersecurity restrictions: OpenAI’s Trusted Access for Cyber program and Anthropic’s Cyber Verification Program. 

These guardrails have been widely criticized, particularly by researchers whose job is to find unknown vulnerabilities in systems and devise ways to exploit them before criminals do.

During a recent appearance on a cybersecurity podcast, Mark Dowd, a well-known security researcher, said that, “it’s not really comfortable to me that these random large companies are making arbitrary decisions about what is safe in security and what’s not.”

Dowd has spent decades finding and selling “zero-days” — previously unknown software flaws and the exploits that take advantage of them — to Western governments, rather than reporting them to the software makers so they get patched. Governments pay a premium for vulnerabilities precisely because they stay open, which is useful for intelligence operations.

Dowd admitted his work may make him biased, but he isn’t alone. Several people who work in offensive cybersecurity — they proactively probe systems for weaknesses — described to TechCrunch how they use AI tools and deal with their guardrails. 

Chris Anley, the chief scientist at security consulting giant NCC Group, said that asking an AI model to try to exploit a bug is a key step in confirming it’s a real vulnerability worth fixing. But if a guardrail prompts the model to refuse to answer the question outright, the guardrail hurts defenders, he said.

“This is where the whole offensive versus defensive and guardrails part comes in, because ‘fix this code’ as a prompt is both an essential mechanism for defense but also a roadmap for finding critical vulnerabilities in the code base,” said Anley. “So at the same time, the same tool is both an offensive tool and a defensive tool, and the two can’t really be unpicked.”

It’s “like a hammer,” he continued. “You can’t build a house without a hammer. It’s definitely a tool but it’s also irreducibly a weapon as well.”

When he and his colleagues run into such a roadblock, they sometimes fall back on open source AI models that come with no guardrails at all.

Paolo Stagno, the chief technology officer at Crowdfense, a well-known company that develops, acquires, and sells unknown vulnerabilities to government agencies, agreed with Dowd, saying AI companies “essentially treat customers like children who need babysitting” with their vetted programs and guardrails. 

Stagno said he and his colleagues do use frontier models — but only for reverse engineering. They avoid using AI to help find vulnerabilities or build exploits, he said, because feeding that work into a cloud-based model risks leaking sensitive vulnerability data or having it absorbed into future training runs. For that step, he said, they use open source models run locally, as they do not rely on sharing data outside of the model. 

Giuseppe Cali, a security researcher who finds zero-days and develops exploits, said guardrails are not impeding his work. That’s because he doesn’t use AI for offensive work; instead, he uses it for initial reverse engineering, to understand the code he’s analyzing, and to build supporting tools. For that, he said, AI tools can speed up the process and allow him to focus on discovering vulnerabilities. 

“I still want to own the actual bug discovery and weaponization myself and that wouldn’t change if all guardrails were lifted tomorrow,” said Cali. “I am jealous of my bugs, and I like this game too much to let models play it for me.”

One researcher at a smartphone-component manufacturer, who spoke on condition of anonymity because he isn’t authorized to talk to the press, said his employer isn’t part of Anthropic’s CVP program and as a result, its tools are barely useful for finding vulnerabilities because the guardrails are too strict.

“If it catches wind we’re doing anything security related, it just stops and isn’t usable,” the person said. 

Chris Thompson — chief executive of cybersecurity firm RemoteThreat and founder of Offensive AI Con, an offensive security and AI-focused event — said that in his experience using the frontier AI models, the guardrails can be inconsistent and work differently every day. That’s true even inside the looser boundaries of Anthropic’s and OpenAI’s vetted programs. 

“I think the practical impact is you spend a lot of time negotiating with the model instead of working on the core security program,” said Thompson. “Instead of analyzing a vulnerability and reasoning through the exploitability, you’re trying to find why you’re getting inconsistent results or why are models over-sanitizing the output.” 

Consequently, researchers rely on or get pushed toward Chinese open source models like GLM — freely downloadable models that can be run locally with no vetting or usage restrictions — said Thompson.

“You have these responsible researchers that are being pushed away from U.S.-governed systems to foreign-owned systems,” he said. “I think it’s more harmful than good to have these guardrails in place.”

Rather than tightening restrictions further, Thompson called for the AI frontier labs to open up their programs, provide responsible access, and hold those who abuse their tools accountable. Otherwise, he argued, defenders will lose the AI race.

“There’s this big storm coming. There’s this big wave of attacks that are going to happen at speed and scale like never before,” said Thompson. “But the same security consulting firms and legit researchers that are trying to make a difference are being stifled right now.”

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

#guardrails #impeding #work #offensive #cybersecurity #researchers #TechCrunchcybersecurity,Zero-days

Post Comment