The Department of Homeland Security bought a dog-like robot that it has modified with an “antenna array” that gives law enforcement the ability to overload people’s home networks in an attempt to disable any internet of things devices they have, according to the transcript of a speech given by a DHS official at a border security conference for cops obtained by 404 Media. The DHS has also built an “Internet of Things” house to train officers on how to raid homes that suspects may have “booby trapped” using smart home devices, the official said.

The robot, called “NEO,” is a modified version of the “Quadruped Unmanned Ground Vehicle (Q-UGV) sold to law enforcement by a company called Ghost Robotics. Benjamine Huffman, the director of DHS’s Federal Law Enforcement Training Centers (FLETC), told police at the 2024 Border Security Expo in Texas that DHS is increasingly worried about criminals setting “booby traps” with internet of things and smart home devices, and that NEO allows DHS to remotely disable the home networks of a home or building law enforcement is raiding. The Border Security Expo is open only to law enforcement and defense contractors. A transcript of Huffman’s speech was obtained by the Electronic Frontier Foundation’s Dave Maass using a Freedom of Information Act request and was shared with 404 Media.

“NEO can enter a potentially dangerous environment to provide video and audio feedback to the officers before entry and allow them to communicate with those in that environment,” Huffman said, according to the transcript. “NEO carries an onboard computer and antenna array that will allow officers the ability to create a ‘denial-of-service’ (DoS) event to disable ‘Internet of Things’ devices that could potentially cause harm while entry is made.” 

DoS attacks are a type of cyber attack where a website, server, or network is overloaded with traffic until it is knocked offline. Huffman did not provide any specifics about how a DoS attack like this would work. But he said DHS wanted to develop this capability after a 2021 incident in which a man suspected of child sexual abuse crimes in Florida used his doorbell camera to see that he was being raided by the FBI and began shooting at them, killing two FBI agents and injuring three others. 

Huffman said that DHS has built something it calls the “FLETC Smart House,” where cops are taught about how smart home devices “could potentially be used against them. They encounter examples of ‘booby traps’ and are shown the evidentiary value of collecting internet of things devices that may contain important evidence related to their case,” Huffman said. 

After the Florida incident, DHS “began researching how the ‘internet of things’ devices may be used to harm law enforcement officers and found numerous postings online showing how these devices can interact with the physical world to create potential ‘booby traps’ for law enforcement. These traps can be used to harm officers and/or destroy evidence,” Huffman said. 

“A suspect who has been searched and is under the control of officers can cause these actions to happen with a simple voice command which can start a chain of events to occur within a house, such as turning off lights, locking doors, activating the HVAC system to introduce chemicals into the environment and cause a fire or explosion to take place,” he added.

The robot is a modified version of Ghost Robotics’ Vision 60 Q-UGV, which the company says it has sold to “25+ National Security Customers” and which is marketed to both law enforcement and the military. “Our goal is to make our Q-UGVs an indispensable tool and continuously push the limits to improve its ability to walk, run, crawl, climb, and eventually swim in complex environments,” the company notes on its website. “Ultimately, our robot is made to keep our warfighters, workers, and K9s out of harm’s way.” 

Correction: An earlier version of this article stated this was a "DDoS" robot, which stands for "distributed denial of service." The speech does not specify whether the denial of service would be "distributed," so we have changed it to the singular "DoS."

One of the worst implementations of Chatbots I’ve seen is Nibble, a chatbot on shopping sites that customers can “negotiate” with on the prices of specific items.

Nibble was founded in 2020, and was covered in some detail in a Forbes story last year, but has drawn a lot of attention this week because of tweet from a user who encountered it while shopping for a mattress that has since been retweeted almost 900 times, and that Nibble cofounder and co-CEO Jamie Ettedgui told me on a call has led to a spike of users interacting with Nibble. 

“This is absolute madness, £80 off for talking to a fucking AI,” Twitter user George McGowan said. Other posts on Twitter said the idea of haggling with a chatbot for a discount was “dystopian,” “Gamification of mattress purchases,” and “the dumbest timeline.”

“I don't think it's dystopian what we're trying to do,” Ettedgui told me. “I know why I get out of bed in the morning, and it’s to not to create some dystopian thing where I'm trying to screw people out of money [...] It comes from my frustration in some ways of being on a website and seeing a voucher code box, and I've got to go hunt around to look for a voucher code that some people have some people don't. I just want to say, look, I've seen your product somewhere else at this price. I would like to buy it here for this price.”

I tried Nibble myself on a couple of sites that use it and found that the chatbot is not actually negotiating with potential customers, but presenting them with a series of minor obstacles before they reach a predetermined maximum discount. Those who don’t have the time to get through those obstacles or don’t understand how the system works will pay more. 

As Ettedgui told me, most retailers that use Nibble will not see their customers hitting the maximum discount Nibble is allowed to give them.

“Let's say the retailer is happy at 15 percent [discount], they set a strategy—preserve margin, a balanced algorithm, and an increased conversion,” Ettedgui said. “Based on all of that the bot is more or less conciliatory when it's negotiating. But even on the most conciliatory, almost nobody will get that 15 percent [discount]. To give you some idea, they're going to be hitting around this sort of 11, 12, 13 percent discount.”

Ettedgui said that if retailers actually want most customers to get that 15 percent discount, they should set the maximum discount or “walk away” price, as Ettedgui called it, at 18 or 17 percent. 

“You have to have a hard coded limit, but that hard coded limit isn't an auto accepted, you don't just get the bottom line,” Ettedgui said. “And loads of people think they're getting the bottom line, and actually I'm sure you saw the tweet and things, and very few of those people are getting the bottom line basically.”

Nibble is a UK-based company, and most of its clients are UK-based as well, though the American shorts brand Chubbies was also a client. To test it, I first headed to Value Lights, “the UK’s leading online lighting retailer.” I saw the Memphis Twist 5 Way Ceiling Light in Black and rather than pay the listed price of £44.99 by adding it to my basket/shopping cart, I clicked the yellow “Negotiate” button below, which opened a window where I could chat with Nibble. 

Off the bat, Nibble said “I could drop the price to 44 just for you, what do you say?”

“I can do 20 for this,” I said. Obviously, that’s an aggressive response, but we’re negotiating, so I came in very low planning to work my way up to a reasonable middle ground. 

Nibble then asked how many of these lights I wanted, to which I responded one unit for 20. 

“Ok, so 1 would come to £44.99 all together. Let’s shake virtual hands at 44?” it asked.

I said 44 was too high, and asked if Nibble could go lower. At that point Nibble declared I got a  £0.99 discount and ended the chat, so I clicked the “Negotiate” button again. When the chat restarted, Nibble made a more aggressive offer.

“I could drop the price to 43.50 just for you, what do you say?” it asked.

Again I offered 20, which Nibble responded to with a crying emoji.

“Not sure we’re on the same page here—or even in the same book! I can’t accept that,” it said. “Would you accept 42?”

I offered 25.

“This is affordable lighting that doesn’t sacrifice style or quality… Talk about a bright Idea!” it said. “Let’s shake virtual hands at 41?”

“30 and you have a deal,” I said, to which it responded with a thinking emoji.

“I can’t accept that, but I've got a good feeling we’re almost there! Let’s wrap this deal at 39.99?”

Overall it went on like this for about 15 minutes, with me moving up slightly in my offers, and Nibble coming in with slightly better discounts. At some point, Nibble took away my ability to type an answer, made me an offer, and allowed me to respond by either clicking “deal” or “no deal” buttons. The virtual version of walking away by closing the chat and restarting it seem to generate the most movement, but I don’t know exactly how Nibble decides when and by how much to discount the price when it does, and the company did not provide more detail when I asked. 

Ultimately, to the best of my ability, and by repeating the test on furniture retailer Bridgman where I tried to buy an outdoor Kensington Single Day Bed, I hit the floor price, maximum discount, or what Ettedgui called the “walkaway price,” and the chatbot wouldn’t budge no matter how much I haggled or how many times I closed and reopened the chat. 

A video on on Nibble’s site explains that it uses generative AI, but that its responses are made up of “atoms,” small pieces of pre-written copy and GPT-generated copy that fit together and put through a quality filter, which altogether limit hallucination risk and fit a retailer’s specific needs. 

Nibble feels in some ways like an AI-powered version of Priceline’s much hyped “Name Your Own Price” system in which customers entered the price they wanted to pay for a flight, hotel, or rental car into a box on a website and waited to see if any hotels, airlines, or rental car companies would sell them what they wanted at the price they wanted. This system was hyped in commercials for years but was ultimately abandoned because consumers didn’t like it. A 2014 study about the Name Your Own Price model found that companies actually make more money when the final price is simply posted as normal. 

I’m a firm believer that if you think the price of a product or service is too expensive, there’s nothing rude about asking for a lower price. The worst that can happen is that the seller will say no, or be offended, at which point I can decide to pay or take my business elsewhere. As Nibble told Forbes last year, the company was founded after Ettedgui enjoyed haggling over the price of a pair of sneakers in a market in Istanbul, a practice he says most people in the UK find impolite. 

Ettedgui pointed me to a TikTok of someone who shared their experience of using Nibble on Chubbies, the shorts brand, saying that it allowed them to negotiate online in a way they wouldn’t have in person. 

“Very often when we do talks and presentations, we [ask] who likes negotiating? In the UK, we tend to get 50/50,” Ettedgui said. “You go anywhere like Dubai, you tend to get about 80 [percent like negotiating] 20 [percent don’t like negotiating]. But when you say do you like negotiating with a chatbot? In the UK, it's like 85 [percent like negotiating] 15 [percent don’t like negotiating] because it's not embarrassing. It's not scary.”

However, my experience of negotiating with Nibble is nothing like my experience of negotiating with people in real human beings. Maybe the person I’m negotiating with is a far better negotiator than me, but a successful negotiation makes me feel like I got a better price, and maybe even built a relationship that extends beyond that specific exchange. It’s a flexible, emotional, and human transaction. Negotiating with Nibble, on the other hand, just feels like I’m forced to jump through a series of arbitrary hoops to reach a predetermined discount that was there to begin with, and that I wish was offered to me without all that effort. At the end of the exchange I either get this optimal discount, or the bot wins, getting me to pay a higher price than I had to.

This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records.

A woman is suing Microsoft and two major U.S. sex toy retailers with claims that their websites are tracking users without their consent, despite promising they wouldn’t do that.

In a complaint filed on June 25 in the Northern District of California, San Francisco resident Stella Tatola claims that Babeland and Good Vibrations—both owned by Barnaby Ltd., LLC—allowed Microsoft to see what visitors to their websites searched for and bought.

“Unbeknownst to Plaintiff and other Barnaby website users, and constituting the ultimate violation of privacy, Barnaby allows an undisclosed third-party, Microsoft, to intercept, read, and utilize for commercial gain consumers’ private information about their sexual practices and preferences, gleaned from their activity on Barnaby’s websites,” the complaint states. “This information includes but is not limited to product searches and purchase initiations, as well as the consumer’s unique Microsoft identifier.” 

The complaint claims that Good Vibrations and Babeland sites have installed trackers using Microsoft’s Clarity software, which does “recording in real time,” and tracks users’ mouse movements, clicks or taps, scrolls, and site navigation. 

Microsoft says on the Clarity site that it “processes a massive amount of anonymous data around user behavior to gain insights and improve machine learning models that power many of our products and services.”

“By allowing undisclosed third party Microsoft to eavesdrop and intercept users’ PPSI in such a manner—including their sexual orientation, preferences, and desires, among other highly sensitive, protected information—Barnaby violates its Privacy Policies, which state it will never share such information with third parties,” the complaint states. 

The complaint includes screenshots of code from the sexual health sites that claims to show  them using Machine Unique Identifier (“MUID”) cookies that “identifies unique web browsers visiting Microsoft sites,” according to Microsoft, and are used for “advertising, site analytics, and other operational purposes.”

The complaint claims that this violates the California Invasion of Privacy Act, the Federal Wiretap Act, and Californians’ reasonable expectation of privacy. 

In February, a woman brought a class action complaint against Adam & Eve, another massive sex toy retailer, claiming that its site gave Google information about her searches for 8-inch dildos and strap-on harnesses. 

Microsoft, Babeland and Good Vibrations did not immediately respond to a request for comment. 

 

A response highlighted the lack of expertise involved in the Cass Review, the lack of evidence for many of the claims, and the lack of any new information that might change the picture of gender-affirming care.

  

by Evan Urquhart

The main professional bodies overseeing the standards for treatment of gender diverse people, WPATH and USPATH, have issued a joint response to the Cass Review, saying they are “extremely concerned” that the Review will serve as a barrier to necessary healthcare for transgender youth. The Review itself, which was published over one month ago, was billed as an independent look at the state of transgender healthcare in England. However, the Review’s objectivity has been questioned due to its asymmetric treatment of the evidence for and against gender-affirming care, resulting in very conservative recommendations that would remove access to medical and even social transition for most patients of the NHS. This is true despite the Review having found little to no evidence of regret, detransition, patients having been rushed into treatment, or any other form of harm stemming mainline treatments for gender dysphoria in youth.

(Curious about what’s in the Cass Review? Read our summary here.)

The primary concern evinced by WPATH/USPATH is for the safety and health of England’s transgender youth. A conservative estimate for the prevalence of trans people in the population is 0.5 percent, meaning that we would expect at least 70,000 young people in that country to grow up to be trans. After an interim report was published by the Cass Review the only available service for gender dysphoria in youth was shut down, and a replacement has yet to be made available.

A recent estimate of the number of patients who have been seen since the GIDS service was shut down is twelve. There are reportedly 5000 young people on the waiting list, still only a small fraction of the number of youth that can be conservatively expected to seek to transition as adults. Despite this, the Cass Review concluded, with no evidence given, that the number of youth seeking treatment was too high to be explained by trans identity alone.

Writing about the current state of trans youth healthcare in England, the WPATH/USPATH statement says, “This is a devastating situation for transgender youth and their families, whose rights are breached as they are being denied medically necessary care. We believe this to be a complete breach of the seven core values enshrined in the NHS Constitution.”

WPATH and USPATH also directly take on the credentials of Dr. Hilary Cass, who was tapped to lead the four-year Cass Review. “Hillary Cass is a pediatrician with hardly any clinical experience or expertise in providing transgender healthcare for young people. Furthermore, Hillary Cass lacks significant research qualifications or research expertise in transgender health,” the statement reads.

Transgender healthcare became controversial in the UK slowly, after a years-long campaign by right-wing tabloids that made the trans community the subject of unceasing negative attention, misinformation, and scare tactics. Although Cass claims in the introduction to the Report that her goal is to improve healthcare for young people (who she proceeds to misgender incessantly throughout the 388-page Report), there is currently no sign of improvement to be had in England, and even trans adults are expressing fears of treatment being withheld, with the specter of forced detransition looming for young and old alike on that foggy, unhappy isle.

---

Evan Urquhart is the founder of Assigned Media and an incoming member of the 2024-2025 Knight Science Journalism fellowship class at MIT.

 

In April, law enforcement contracting giant Axon announced Draft One, a new product that uses OpenAI’s GPT-4 Turbo to automatically generate police reports from officers’ bodycam audio. The product promises to make police work more efficient, freeing up resources for officers to direct elsewhere, while also raising serious questions about bias, accuracy, and the role of AI in policing. 

Since then, Axon has held webinars with prosecutors and law enforcement officials around the country, according to emails obtained by 404 Media and a review of recordings of the webinars online. Those recordings include some of the concerns and hopes of prosecutors around the technology, as well as show Draft One in action, providing more insight on how exactly it works.

“I’d have to see many reports to understand that the AI was capable of writing a cogent, relatively accurate report,” Bryan Porter, Commonwealth’s Attorney for the City of Alexandria, VA, said in one of the webinars. He adds he would need lawyers who are working on cases to review artificially generated reports to make sure “that they are on point and helpful in the prosecution of their cases.”

On the day of Draft One’s announcement, Axon emailed law enforcement officials about the product launch and provided details about the upcoming webinar called “AI in Criminal Justice: What Prosecutors Need to Know,” according to a copy of an email sent to the San Diego Police. 404 Media obtained the email and others through a public records act request.

“We know that this kind of new AI technology will impact the criminal justice process, and so we wanted you to know about it early and give you the chance to ask questions,” the email reads. “We’ve spoken with a number of prosecutors across the United States already, and you’ll even hear from three different prosecutors during this webinar.”

Included in the webinar is a video demonstration of Draft One. It starts with Noah Spitzer-Williams, senior principal, product management at Axon, showing a video filmed with an Axon bodycam. In that, a man called Marcus tells a police officer about a suspect who was yelling at a family, and provides a description of the suspect. Axon told 404 Media in an email that this incident was not genuine, and filmed for demonstration purposes.

Spitzer-Williams then opens up Draft One. The tool has a white, grey, and blue interface, with an option to start dictation, or click a button on the right which says “generate a draft narrative from the audio recordings of your incident.” He clicks that button, navigates to a search tab, and selects the previously shown video. 

The system then asks the user to select what type of incident they are making a narrative from. A multiple choice question has boxes for “domestic dispute,” “sexual offense,” “drug related,” “fraud and financial,” and many more. Spitzer-Williams says this selection impacts the generated output, because “we know that an impaired driving incident looks very different than a simple theft, for example.”

The user also needs to select the charge severity of the incident, be that no charge, an infraction, a misdemeanor, or a felony. This is done, Spitzer-Williams says, so customers can restrict Draft One’s use based on the incident type or the level charges. (At an Axon event recently, Spitzer-Williams claimed that most agencies using the tool are already using it for all incident types).

Finally, Spitzer-Williams clicks “generate draft.” The screen then tells officers to proofread the draft, make any corrections, and sign off on its accuracy. A few seconds later, Draft One produces the report. It includes sections where the officer is asked to provide additional information or context. 

“Marcus described the suspect as wearing a green jacket with the hood up and blue jeans, approximately the same height as himself. [INSERT officer’s observation of Marcus’s height for reference],” the report reads. In the demonstration, the system also deliberately inserted some errors that the officer must manually remove before the report can be submitted; this is to ensure that the officer is proofreading the report, Spitzer-Williams said. In this case, one of the mistakes was that “an emotional support alligator was stolen,” which did not happen in the bodycam footage.

After making their changes, the officer has to sign an acknowledgment that they used Draft One to generate the report, but that they are willing to testify to the accuracy of the report.

Matthew Guariglia, senior policy analyst at activist organization the Electronic Frontier Foundation (EFF), told 404 Media in a statement that “Policing, with its legal power to kill, detain, or ultimately deny people’s freedom, is too powerful an institution to outsource its official narratives to technologies in a way that makes officers immune to critique, transparency, or accountability because they have theoretical plausible deniability that they did not review the report thoroughly and therefore were not aware of mistakes the AI made.”

“We still have a tremendous number of questions about how the technology would reflect action, spoken words, and introduce uncertainty into reports—for instance, an officer shouting ‘stop resisting’ and whether it would report it as the officer's command or as the hypothetical resistance as fact. For that reason, we believe municipalities with the ability to control police purchases should show extreme restraint when it comes to deploying this new and problematic technology,” he added.

Axon told 404 Media in a statement that “While Draft One provides several benefits to law enforcement, we also believe it will benefit prosecutors who work with police agencies that leverage Draft One. Feedback from trial agencies indicate that prosecutor offices will receive reports that are more consistent, clear and detailed, which can help accelerate the justice process by delivering higher quality reports, faster.” The company added that “With Draft One, police narrative reports continue to be the responsibility of officers. Critical safeguards require every report to be edited, reviewed and approved by a human officer, ensuring accuracy and accountability of the information.”

A team of scientists in Australia say that they have found a way to make cold brew coffee in less than three minutes using an ultrasonic reactor. This is a potentially massive deal because cold brew normally takes between 12 and 24 hours to brew, a problem for me, personally, when I do not carefully manage my cold brew stock. The lead scientist on the research team tells me he has also created a “cold espresso,” which is his personal favorite and sounds very intriguing.

The researchers at the University of New South Wales Sydney claim that their ultrasonic extraction held up to a “sensory analysis” and blind taste tests by trained experts: “A sensory analysis was conducted to evaluate appearance, aroma, texture, flavor, and aftertaste, which demonstrated that coffee brewed for 1 and 3 min in the sonoreactor exhibited almost undistinguishable properties compared to a standard 24 hour [cold] brewing without ultrasound,” they write in a paper about the method in the journal Ultrasonics Sonochemistry. 

For the uninitiated, cold brewed coffee is made by soaking coffee grounds in cold or room temperature water in large batches to create a concentrate that you can keep in the fridge for a week or two. Because the water is not hot, the extraction from ground coffee beans takes much longer than it does with traditional hot brewing. The resulting cold brew is less acidic, less bitter, and sweeter. This long brew time isn’t a problem if you plan ahead, but, as mentioned, if you do not plan ahead, you cannot really speed up the cold brew time while continuing to have cold brew. As lead author Francisco Trujillo notes in the paper, the resulting large batches of cold brew concentrate also take up a lot of counter and fridge space, meaning that not every coffee shop or restaurant has it on hand. This is a phenomenon I am very familiar with, as many establishments currently on my shitlist claim that they have “cold brew” that is actually hot coffee poured over ice. 

Trujillo’s new method uses room temperature water in a normal espresso machine that has been modified to turn the boiler off (or down) and has been modified to add a device that hits the beans with ultrasonic waves at a specific frequency that makes the beans shake. In layman’s terms, they are blasting the beans with ultrasounds, which cause the beans to vibrate and its cell walls to burst, allowing the rapid extraction of coffee without heat. Trujillo explains in the paper that extraction happens  because of “acoustic cavitation. When acoustic bubbles, also called inertial bubbles, collapse near solid materials, such as coffee grounds, they generate micro jets with the force to fracture the cell walls of plant tissues, intensifying the extraction of the intracellular content.” 

Trujillo told me that he learned this was possible in a study he published in 2020, and set to “superimpose ultrasound in the coffee basket of an existing espresso machine. We purchased a few Breville espresso machines, opened them up, and started the journey. Mathematical modeling of the sound transmission system and of acoustic cavitation was key for the success of the design.” Some of that mathematic modeling is available in the paper here:

He said that they experimented with a variety of different frequencies, and said that frequencies between 20-100 kHz are all good at extracting coffee. “The lower the frequency, the larger the transducer and the horn,” he said. “If the frequency is in the low range, there are harmonics that can be heard. We worked at 28 kHz and at 38-40 kHz, and we chose 38-40 kHz as it was more compact and with a quieter performance.” 

Essentially, his team was able to modify an existing Breville espresso machine to do this, and said that they experimented with different brew times, and water temperatures (104 degrees F, well below boiling, was the hottest they tried) and were able to create a variety of different cold extractions, including one that is not mentioned in the paper but which Trujillo told me about that he calls “cold espresso” and which he said are his "favorite ones" and “offer a unique sensory experience like nothing in the market. It is bold and rich in flavor and aroma, less bitter, and with a great acidity. It is more viscous and with a very nice finishing (according to coffee experts that have tried our brews).  That will be a unique and novel coffee beverage to be appreciated by coffee lovers, hopefully worldwide.” 

The various ultrasonic cold brews the team produced were tested by a team at the Queensland Alliance for Agriculture and Food Innovation by a group of “11 trained sensory panelists” who “had previously been screened of their sensory acuity.” They scored the ultrasonic extractions very similarly to real cold brew, though of course whether the ultrasonic coffee is actually “almost undistinguishable” from real cold brew will depend on each person’s taste.

I have long been interested in the science of coffee. When I was a freelancer, I went to Manizales, Colombia, to a national laboratory called “Cenicafe.” A scientist there called it the “NASA of Colombia,” referring to how seriously the institute takes the scientific pursuit of growing, roasting, and brewing ever-improving coffee. Cenicafe was easily one of the coolest places I’ve been in my life; they were genetically sequencing different species of coffee, hybridizing arabica and robusta coffee in attempts to create strands that taste good but are also resistant to both climate change and “coffee rust,” a fungus that regularly blights huge amounts of the coffee harvest in many countries, and were experimenting with new ways to brew coffee. I include this to say that, while inventing a new type of coffee brewing may seem frivolous, there is actually a huge amount of time, effort, and funding going into ensuring that there is ongoing innovation in coffee growing and brewing tech, which is particularly important considering that coffee plants are particularly susceptible to climate change.

Trujillo said that he plans to license the technology to coffee maker companies so that it can be used in both commercial coffee shops and in people’s homes. 

“I love cold brew, and coffee in general,” he said. “I am Colombian and my grandfather had a business of buying coffee beans from the local producers, he then dried the beans under the sun on ‘costales’ (a traditional Colombian strong fabric) that he placed on the street. That was in Ortega, a little town in Colombia. There were other gentlemen like my grandfather who had the same business. So, during the season period, the streets of Ortega were filled with costales with coffee beans drying under the sun!”