As Large Language Models (LLMs) are increasingly deployed in sensitive domains such as enterprise and government, ensuring that they adhere to user-defined security policies within context is critical-especially with respect to information non-disclosure. While prior LLM studies have focused on general safety and socially sensitive data, large-scale benchmarks for contextual security preservation against attacks remain lacking. To address this, we introduce a novel large-scale benchmark dataset, CoPriva, evaluating LLM adherence to contextual non-disclosure policies in question answering. Derived from realistic contexts, our dataset includes explicit policies and queries designed as direct and challenging indirect attacks seeking prohibited information. We evaluate 10 LLMs on our benchmark and reveal a significant vulnerability: many models violate user-defined policies and leak sensitive information. This failure is particularly severe against indirect attacks, highlighting a critical gap in current LLM safety alignment for sensitive applications. Our analysis reveals that while models can often identify the correct answer to a query, they struggle to incorporate policy constraints during generation. In contrast, they exhibit a partial ability to revise outputs when explicitly prompted. Our findings underscore the urgent need for more robust methods to guarantee contextual security.

Throughout the unprecedented bombing campaign that has defined Israel's genocidal war on Gaza, Palestinians there have lived with a near constant, inescapable sound of drones. It's a sound that signals death could be close. Hind Hassan tracks how the Israeli military has dramatically increased its use of drones and artificial intelligence (AI) to surveil, track and kill Palestinians. In Gaza, this technology has produced a kill rate higher than any other 21st-century conflict. But its implications are far greater – creating the potential for armies of the future to inflict maximum destruction on their targets with minimal risk to themselves.

Industries can rethink how work gets done, raising the bar for productivity and workplace safety. Healthcare systems worldwide are struggling with overcrowded hospitals, physician burnout, and rising surgery delays. Which is why it's always a good thing to see research exploring new solutions through technology. The University of California San Diego (UCSD) is looking into humanoid robots as a potential solution. It suggests they might play a vital role in easing medical burdens.

Achieving that perfectly manicured lawn is a whole lot less time-consuming now that a robot can do the job for you. And over the long term, a robot lawn mower will cost less than hiring a landscaper--and it won't expect a tip every week. Top-of-the-line models cost a pretty penny, but they can climb slopes and handle very large yards (we're talking acres of grass). While early robot lawn mowers required you to lay down a boundary wire to prevent them from wandering out of your yard, each of mowers listed here use advanced navigation technology that eliminates the need for any wires. The good news is that we've already spotted some great deals on robot lawn mowers ahead of Amazon's Prime Day sale, and we expect there will be more when the sale actually gets underway on July 8.

Wybot's solar skimmer does a surprisingly good job of grabbing leaves off the surface of the pool, but its loud operation and poor power management knock it down a peg. Solar-powered pool skimmers flit along the surface of your pool operating under the idea that if they can scoop up debris before it sinks, you won't need to clean the bottom of the pool. It sounds logical, but in practice, most pool skimmers don't do the absolute best of jobs--there's only so much surface area a skimmer can cover before leaves get waterlogged and sink to the depths. But robotic skimmers are better than nothing, especially if you don't have a good in-wall skimmer. The Wybot F1 Pool Skimmer was much more effective at capturing floating leaves than any skimmer I've used to date.

With pool-mapping capabilities and other smart features, the Dreame Z1 Pro is one of the most intelligent robots I've tested to date. From the start, Dreame's Z1 Pro robotic pool cleaner certainly seems to check off all the boxes. Its features list touts just about everything: The ability to clean floor, walls, and waterline. I'm not sure what the touted "Triple Surround Fusion Perception System" is, but that sounds good, too. I'll start with what I liked the most: After unboxing, I discovered that the 27-pound robot offers one of the most convenient charging systems I've seen to date, thanks to a magnetic charging mechanism that simply snaps onto the device's chassis, with no plugs or rubber gaskets involved--and no need to hoist the robot onto a bulky charging station, either.

Breakthroughs, discoveries, and DIY tips sent every weekday. Our smartphones have transformed the way we take photos and videos and our relationship to these digital memories. Most of us will snap at least some pictures and clips every day with the gadget that's always close at hand. If you want to get more creative with photos on your phone, you can. Sometimes you're going to want to take a picture remotely, without your phone in your hand and your finger over the shutter button--maybe you're taking a wide shot of a large group, or you want to capture a lot of your surroundings.

The Fanttik Aero X robotic pool cleaner runs fast and runs long: With six hours of running time and top-notch cleaning power, the device makes short work of underwater debris. In a world dominated by bulbous black-and-blue hardware, the Fanttik Aero X pool robot immediately caught my eye. It's not just that it's black and yellow, it's that the industrial design looks more like a lawn mower than any pool robot I've tested. It has much smaller front wheels than usual, and an exposed rubber drive belt that connects them to its motor. The forward-center brush cylinder is even reminiscent of the front of a lawn mower.

Language-conditioned robot models (i.e., robotic foundation models) enable robots to perform a wide range of tasks based on natural language instructions. Despite strong performance on existing benchmarks, evaluating the safety and effectiveness of these models is challenging due to the complexity of testing all possible language variations. Current benchmarks have two key limitations: they rely on a limited set of human-generated instructions, missing many challenging cases, and they focus only on task performance without assessing safety, such as avoiding damage. To address these gaps, we introduce Embodied Red Teaming (ERT), a new evaluation method that generates diverse and challenging instructions to test these models. ERT uses automated red teaming techniques with Vision Language Models (VLMs) to create contextually grounded, difficult instructions. Experimental results show that state-of-the-art models frequently fail or behave unsafely on ERT tests, underscoring the shortcomings of current benchmarks in evaluating real-world performance and safety. Code and videos are available at: https://sites.google.com/view/embodiedredteam.

We may earn revenue from the products available on this page and participate in affiliate programs. If you're tired of stockpiling blankets, extra socks, and heated slippers to keep you warm, it might be time to consider getting a space heater. These powerful appliances are a great way to get cozy without installing a complicated heating system or commandeering the thermostat. If your radiator just isn't cutting it or someone insists on keeping a window open to freshen the room up, a space heater could be the perfect solution. These hot machines are designed specifically to warm up spaces of all sizes and should be portable, effective, and fast-acting. Our best overall pick, the Lasko 5586 Electric 1500W Ceramic Space Heater Tower, ticks all these boxes.