Tag: technology

Invisible text that AI chatbots understand and humans can’t? Yep, it’s a thing.

What if there was a way to sneak malicious instructions into Claude, Copilot, or other top-name AI chatbots and get confidential data out of them by using characters large language models can recognize and their human users can’t? As it turns out, there was—and in some cases still is.

The invisible characters, the result of a quirk in the Unicode text encoding standard, create an ideal covert channel that can make it easier for attackers to conceal malicious payloads fed into an LLM. The hidden text can similarly obfuscate the exfiltration of passwords, financial information, or other secrets out of the same AI-powered bots. Because the hidden text can be combined with normal text, users can unwittingly paste it into prompts. The secret content can also be appended to visible text in chatbot output.

The result is a steganographic framework built into the most widely used text encoding channel.

“Mind-blowing”

“The fact that GPT 4.0 and Claude Opus were able to really understand those invisible tags was really mind-blowing to me and made the whole AI security space much more interesting,” Joseph Thacker, an independent researcher and AI engineer at Appomni, said in an interview. “The idea that they can be completely invisible in all browsers but still readable by large language models makes [attacks] much more feasible in just about every area.”

To demonstrate the utility of “ASCII smuggling”—the term used to describe the embedding of invisible characters mirroring those contained in the American Standard Code for Information Interchange—researcher and term creator Johann Rehberger created two proof-of-concept (POC) attacks earlier this year that used the technique in hacks against Microsoft 365 Copilot. The service allows Microsoft users to use Copilot to process emails, documents, or any other content connected to their accounts. Both attacks searched a user’s inbox for sensitive secrets—in one case, sales figures and, in the other, a one-time passcode.

When found, the attacks induced Copilot to express the secrets in invisible characters and append them to a URL, along with instructions for the user to visit the link. Because the confidential information isn’t visible, the link appeared benign, so many users would see little reason not to click on it as instructed by Copilot. And with that, the invisible string of non-renderable characters covertly conveyed the secret messages inside to Rehberger’s server. Microsoft introduced mitigations for the attack several months after Rehberger privately reported it. The POCs are nonetheless enlightening.

Keep reading

Open AI Wants to Build Data Centres That Would Consume More Electricity Per Year Than the Whole of the U.K.

Over the past few months, the newswires have been hot with stories about the large-scale data centres that will be required to meet the needs of the forthcoming revolution in Artificial Intelligence (AI). How much electricity will these new data centres consume and what does that mean for the electricity demand forecasts underpinning the plans for Net Zero?

Recent Date Centre Announcements

To give a flavour of the scale of data centre developments that are coming, it is helpful to look at recent announcements from large tech companies. Back in March, it was announced that Amazon had bought a 960MW data centre that is powered by an adjacent nuclear power station. In April, Mark Zuckerberg CEO of Meta that owns Facebook and Instagram said energy requirements may hold back the build out of AI data centres. He also talked about building data centres that would consume 1GW of power.

Last month, Oracle chairman Larry Ellison announced that Oracle was designing a data centre that would consume more than 1GW that would be powered by three small modular nuclear reactors (SMRs). Then Microsoft also got in on the act when it announced it had done a deal with U.S. utility Constellation to restart the 835MW Three Mile Island (TMI) Unit 1 nuclear power plant to power its data centres. Anxious not to be left out, Sundar Pichai, CEO of Google said they too were working on 1GW data centres and saw money being invested in SMRs.

Finally, Sam Altman of OpenAI, the creator of ChatGPT has trumped them all by pitching the idea of 5GW data centres to the White House. Altman has been heard talking of building five to seven of these leviathans.

Bloomberg, usually a driver of the Net Zero band wagon, has reluctantly admitted that there’s not enough clean energy – nuclear or otherwise – to satisfy AI’s voracious appetite and gas will have to fill the gap. America’s national security and energy security is eclipsing climate concerns.

This admission is important because it acknowledges that such important infrastructure cannot rely upon the vicissitudes of the weather. These data centres need a source of electricity that is reliable and always available. It would be absurd to think that a ChatGPT user will want to receive a message saying their request cannot be processed because it is not windy enough.

Keep reading

Swedish Police Want to Fight Crime with Live Facial Recognition

The Swedish police want to use facial recognition in real time to crack down on serious crimes.

Government investigators have already drafted a bill that will make it possible to use the technology. The regulation, however, still needs to be completed before it can be tabled, National Police Chief Petra Lundh told publicly funded radio broadcaster Sveriges Radio last week.

Lundh also noted that the legislation must comply with the EU AI Act and could potentially be temporary until crime rates settle down.

Sweden has been experiencing a flood of gang-related attacks, including firearms and explosives, leading the Scandinavian country to crown itself with the title of highest per capita gun violence rate in the European Union. Police Chief Lundh believes law enforcement agencies could use cameras to find suspects.

“It is not unusual that we have a picture of the likely perpetrator, but then we cannot find him or her,” Lundh says.

The suggestion has already been met with criticism. The technology could make incorrect matches for people with dark skin leading to perceptions that the AI is racist, says lawyer Kristofer Stahre.

“I am worried about what consequences it may have for the Swedish people,” he says.

The Swedish government has been working on expanding the use of biometric data in policing on other fronts.

Keep reading

In a rare disclosure, the Pentagon provides an update on the X-37B spaceplane

After more than nine months in an unusual, highly elliptical orbit, the US military’s X-37B spaceplane will soon begin dipping its wings into Earth’s atmosphere to lower its altitude before eventually coming back to Earth for a runway landing, the Space Force said Thursday.

The aerobraking maneuvers will use a series of passes through the uppermost fringes of the atmosphere to gradually reduce its speed with aerodynamic drag while expending minimal fuel. In orbital mechanics, this reduction in velocity will bring the apogee, or high point, of the X-37B’s orbit closer to Earth.

Bleeding energy

The Space Force called the aerobraking a “novel space maneuver” and said its purpose was to allow the X-37B to “safely dispose of its service module components in accordance with recognized standards for space debris mitigation.”

While the reusable Boeing-built X-37B spaceplane is designed to land like an aircraft on a runway, the service module, mounted to the rear of the vehicle, carries additional payloads. At the end of the mission, the X-37B jettisons the disposable service module before reentry. The Space Force doesn’t want this section of the spacecraft to remain in its current high-altitude orbit and become a piece of space junk.

“Once the aerobrake maneuver is complete, the X-37B will resume its test and experimentation objectives until they are accomplished, at which time the vehicle will deorbit and execute a safe return as it has during its six previous missions,” the Space Force said.

The Space Force has identified mobility in orbit as a key focus for its next-generation space missions. This would allow satellites to more freely move between altitudes and orbital inclinations than they can today. Commanders don’t want a spacecraft’s movements to be constrained by the amount of fuel it carries, allowing satellites to “maneuver without regret.”

Space Force leaders have discussed in-orbit refueling, more efficient propulsion technologies, and other ways to achieve this end. Aerobraking is another way to lower a spacecraft’s orbit without using precious propellant.

“This first-of-a-kind maneuver from the X-37B is an incredibly important milestone for the United States Space Force as we seek to expand our aptitude and ability to perform in this challenging domain,” said Gen. Chance Saltzman, the Space Force’s chief of space operations.

Keep reading

The U.S. Government Is Dramatically Expanding The Use Of Facial Recognition Technology

Do you want to live in a society where you are required to have your face scanned wherever you go?  If not, you may want to speak up now while you still can.  As you will see below, the U.S. government is aggressively expanding the use of facial recognition technology for identification verification purposes.  For now, the use of facial recognition technology will be optional.  But as we have seen before, once a voluntary option is adopted by enough people our leaders have a way of making it mandatory.  Of course it isn’t just our government that is pushing facial recognition technology.  It is popping up throughout our society, and given enough time it would literally be everywhere.

Login.gov is billed as “a single sign-on solution for US government websites”, and now users of Login.gov will be given the option to use facial recognition technology to verify their identities…

An online hub for Americans to access benefits and services across the federal government is giving its users a new option to sign on.

The General Services Administration will begin offering facial recognition technology as an option for users of Login.gov, a one-stop for government-provided public services, to verify their identities.

GSA’s Technology Transformation Services announced Wednesday it will allow Login.gov users to verify their identity online through facial technology that meets standards set by the National Institute of Standards and Technology’s 800-63-3 Identity Assurance Level 2 (IAL2) guidelines.

We are being told that this will help reduce identity theft and fraud, and I don’t know anyone that likes identity theft and fraud.

But do we really want to live in a dystopian world where our faces are constantly being scanned all the time?

I certainly don’t.

Keep reading

Ancient Construction Technology Discovered Under a Neolithic House

A groundbreaking discovery in Denmark has revealed evidence of an advanced technological structure dating back 5,000 years. While excavating a Neolithic site on the island of Falster, archaeologists uncovered a stone-paved root cellar for storing produce beneath an ancient dwelling. This significant find has challenged existing understandings of Neolithic life in Scandinavia, where early agricultural communities were thought to have employed simpler preservation methods.

The excavation, led by researchers from the Museum Lolland-Falster and Aarhus University, has been documented in a detailed study published in Radiocarbon.

The site at Nygårdsvej 3, was uncovered during construction work for a railway. It has proven to be an archaeological gold mine, well, as regards ancient architecture is concerned. Archaeologists identified two phases of house construction, both attributed to the Funnel Beaker Culture (also known as TRB or Trichterbecherkultur). This culture, which emerged around 4000 BC, marked the region’s shift from a hunter-gatherer society to a more sedentary lifestyle centered on agriculture and animal husbandry.

The houses discovered at the site followed a common architectural design of the period, known as the Mossby-type, which featured large double-span roofs supported by posts. The first house phase, made between 3080 and 2780 BC, had 38 postholes, while the second phase contained 35.

The floors of the houses were made from compacted loam, a mixture of sand and clay that provided a durable and stable surface. This is a construction material still in use in various parts of the world today.

Keep reading

Protective Nets To Shield F-22s Eyed For Airbase Swarmed By Mystery Drones

U.S. Air Force officials at Langley Air Force Base in Virginia are looking at installing anti-drone nets to help protect F-22 Raptor stealth fighters on the flightline. This comes nearly a year after the base was subjected to waves of still-mysterious drone incursions, which The War Zone was first to report. It also underscores the U.S. military’s continued lag when it comes to responding to the very real threats posed by uncrewed aerial systems, at home and aboard, and particular hurdles to doing so domestically.

Langley’s 633rd Contracting Squadron put out a notice on October 4 asking for information about potential counter-drone netting that could be installed around up to 42 existing open-ended sunshade-type shelters at the base. Langley, now technically part of Joint Base Langley-Eustis, is one of a select few bases to host F-22s and is a key component of the Air Force’s posture to defend the U.S. homeland.

The 633rd “is in the process of determining the acquisition strategy to obtain non-personal services for the Unmanned Ariel Services (UAS) Netting for East Ramp Metal Sunshades,” according to the contracting notice. “The intention of the netting is to deter and ultimately prevent the intrusion of UAS’s near airmen and aircraft. This initial sunshade netting installation on the metal sunshade (bay Alpha 1) shall serve as a proof of concept for the remaining sunshades.”

The “netting should be capable of disabling a Group 1/ “Small” Unmanned Aerial Systems (UAS), such as the DJI Matrice 300 RTK, while remaining attached,” the notice explains. Per the U.S. military’s definitions, drones in Group 1 can have weights of up to 20 pounds, fly up to 1,200 feet, and reach speeds of up to 100 knots.

Keep reading

Lucid Dreaming Breakthrough: Startup Claims First-Ever Two-Way Dream Communication

In a development that sounds straight out of science fiction, REMspace, a California and Russia-based neurotech startup, claims to have achieved the first two-way communication between individuals during lucid dreaming.

Using specially designed equipment, participants reportedly exchanged a message while asleep—an extraordinary claim that has yet to be peer-reviewed.

This milestone, if validated, could mark a turning point in dream research, with REMspace suggesting applications from mental health therapies to skill training.

Communication While Dreaming

REMspace is a neurotechnology company specializing in sleep enhancement and lucid dreaming. Using specially designed equipment, REMspace claims that two individuals successfully induced lucid dreams and exchanged a simple message with each other.

In May 2023, REMspace founder Michael Raduga made headlines after reportedly drilling into his own skull to implant a microchip in an attempt to control his dreams. Raduga, who shared details and graphic images on social media, claims the chip was designed to stimulate his brain’s motor cortex during REM sleep. Despite nearly dying from blood loss, he remains optimistic about the experiment’s potential.

It is important to highlight that while Raduga describes his self-administered procedure as groundbreaking, Raduga is not a qualified neurosurgeon.

Lucid dreaming, according to WebMD, is the state of being aware that you are dreaming while asleep. While around 50 percent of people report experiencing at least one lucid dream, the idea of communication within such a state is still in its early stages of research.

Keep reading

How the Army is using AI during Hurricane Helene relief

The Army’s 18th Airborne Corps is for the first time using a battlefield capability to map road closures, cellular outages, supply needs and other data in real time to help the Federal Emergency Management Agency and U.S. Northern Command help people whose homes and communities were battered by Hurricane Helene late last month.

The Army is using its Maven Smart System to provide responders with the information needed to make quick, on-the-ground decisions, such as where to send medical supplies or how many truckloads of water to take into certain storm-ravaged areas, defense officials told reporters Monday.

Weeks after the deadly hurricane tore a path from Florida’s Gulf Coast into the Appalachian Mountains, some residents in the southeast are still sifting through the wreckage caused by floods and landslides that destroyed entire towns.

More damage is feared as Hurricane Milton bears down on Florida this week as well.

Keep reading

Vigilante Swiss teens use dating apps to lure paedophiles into violent ambushes – before being caught and charged over their retribution

A group of vigilante teenagers took the law into their own hands and used dating apps to lure paedophiles into violent ambushes after police ignored their complaints about a schoolgirl pal who was being groomed. 

The schoolchildren, aged between 13 and 18, from Lugano in Switzerland were themselves apprehended by police earlier this month.

They were held on charges including grievous bodily harm, assault, coercion, robbery, false imprisonment, and extortion.

The elaborate scheme was the brainchild of a 13-year-old boy who hatched a plan to use dating apps, such as Tinder, to track down adults who were trying to meet minors.

It was set up after local police allegedly ignored their complaints about the harassment of an underage girl by a man who was sending her nude pictures. 

Conversations between the two were even shared with no result, so the children came up with their own solution.

After chatting to the adults on dating apps, plans would be made over Whatsapp or Instagram to meet in-person at parks or even in flats across the Swiss city.

Once there, the unsuspecting targets walked straight into the group’s carefully laid trap.

The alleged paedophiles would first be greeted by an underage girl or boy, whose role was to persuade them to undress.

Once this was accomplished, the group of teens would converge on the scene, kicking and punching the individual while simultaneously urinating on them, spitting at them, or shaving off their hair.

The actions were reportedly recorded and sometimes shared with third parties. The group had even considered broadcasting the acts live on social media.

When interviewed by police, one of the teens said: ‘It all started when a 35-year-old man started harassing an underage friend of mine by sending her nude photos and asking her for sex.

‘We tried to file a complaint, but we were not taken seriously, we even showed the officials the chats.’

Keep reading