A vital segment of the internet’s often unseen infrastructure experienced a worldwide outage on Tuesday, leading to error messages appearing across various websites.
Cloudflare, a US-based firm that safeguards millions of websites against malicious assaults, faced an unexplained problem that hindered internet users from accessing certain client sites.
Some website owners struggled to reach their performance dashboards. The company reported that platforms like X and OpenAI saw a spike in outages concurrent with the Cloudflare issue. Down Detector.
The outage was first noted at 11:48 a.m. London time, and by 2:48 p.m., Cloudflare announced: “A fix has been implemented and we believe the issue is resolved. We are continuing to monitor the situation to ensure all services return to normal.”
A spokesperson from Cloudflare issued an apology “to our customers and the entire internet for the disruptions today.” They added, “We aim to learn from today’s events and enhance our services.”
To tackle the issue, the company turned off an encryption service known as Warp in London, stating that “users in London attempting to access the internet via Warp will face connection failures.”
Professor Alan Woodward of the Surrey Cyber Security Centre referred to Cloudflare as “the biggest company you’ve never heard of.” The firm claims to provide services that “protect websites, apps, APIs, and AI workloads while enhancing performance.”
Woodward characterized Cloudflare as a “gatekeeper,” noting its role in monitoring site traffic to protect against distributed denial-of-service (DDoS) attacks, which involve malicious actors attempting to overwhelm a site with requests. It also verifies whether a user is human.
Upon identifying the fix, Cloudflare revealed that the issue stemmed from “configuration files automatically generated to manage threat traffic.”
These files expanded larger than anticipated, resulting in a failure of the software systems that direct traffic for numerous Cloudflare services.
“To clarify, there is no evidence that this was caused by an attack or malicious activity,” the spokesperson stated. “We anticipate that some Cloudflare services may experience temporary degradation as traffic spikes following this incident, but we expect all services to normalize within the next few hours.”
This incident at Cloudflare occurred less than a month after an outage at Amazon Web Services disrupted thousands of sites.
According to Woodward, “We’re beginning to realize just how few companies are integral to the internet’s infrastructure, so when one fails, the impact is immediately noticeable.”
Although the cause remains unidentified, Woodward suggested that it’s unlikely to be a cyber attack, as such a major service typically does not have a single point of failure.
Over 1,100 individuals have faced charges related to the summer 2024 riots, with a small fraction being prosecuted for crimes associated with their online conduct.
Sentences varied from 12 weeks to seven years, igniting a surge of online backlash. The individuals behind the posts were varied; one notable case is that of I defended, who emerged as a cause célèbre and was labeled a “political prisoner.” Their posts were minimized and mischaracterized; their prosecution was framed as an infringement on free speech, despite the majority of online-related charges involving allegations of inciting racial hatred.
The posts did not predominantly surface in mainstream social media platforms like X, Instagram, or Facebook, but rather in niche online spaces commonly linked to fringe ideologies like Telegram, Parler, GetTr, 4Chan, and 8Kun. While many of these posts were on personal profiles, some appeared in public group forums.
This raised questions: What online communities did these individuals engage with, and who were their advocates? What type of content was circulating in these environments? It seemed that within these circles, views were so normalized that individuals felt emboldened to share content that was considered criminal by British authorities and the judiciary.
As a starting point, we utilized publicly accessible resources (police records and news reports) to track Facebook accounts of those implicated in previous investigations. Out of approximately 20 individuals charged with online offenses related to the summer 2024 riots, we followed five to three public Facebook groups. We also discovered visually similar or replicated posts defending those referenced in these groups.
This led to the mapping of a broader network of other Facebook groups, connected through shared memberships and group moderators and administrators.
In this exploration, we uncovered vibrant ecosystems characterized by a profound distrust of government and its institutions, alongside online communities preoccupied with anti-immigrant sentiments, naturalism, conspiracy theories, and misinformation.
Additionally, we found individuals who expressed genuine concerns about the society they belong to, alongside those who are deeply disillusioned and believe their freedom of expression is at risk.
Identification of Groups
Why focus on these groups?
Three groups were selected for the primary analysis because they included one or more current or former members charged in connection with the summer 2024 riots, or individuals involved in the riots who made comments either in person or online.
We established links between these and 13 additional groups, with all but three being public. These groups play significant roles, as moderators can oversee memberships, approve requests, and issue bans, with the authority to delete posts and comments. Administrators have even broader permissions, including the ability to modify group settings, update descriptions, and appoint additional moderators or administrators.
Which posts were analyzed?
To understand the type of content shared within these groups, we aimed to capture all posts made by the three largest groups from their inception until mid-May 2025.
We collected links and text from a total of 123,000 posts. However, due to the classification process (outlined below), the analysis was ultimately focused on 51,000 text-based posts.
What was the group membership size?
We did not record the names of individual group members (aside from moderators, administrators, and prominent posters). Therefore, when discussing combined memberships across groups, it is likely that individuals who belong to multiple groups were counted more than once.
Classification
First, we verified that the posts contained far-right content using established academic methods and categorized them through specific keywords indicating radicalization. We supplemented this with an AI tool that became available to data teams due to recent changes in editorial policies surrounding its journalistic use, classifying content as anti-establishment, anti-immigrant, migrant demonization, naturalism, and far-right identity/denial.
For categorizing the 51,000 social media posts, we employed ChatGPT 4.1 via OpenAI’s API. The prompts underwent rigorous testing across a random sample of 12 iterations, ensuring that at least two reviewers concurred until a consensus of over 90% was reached between the model and three human reviewers.
We are confident in the model’s reliability in small batches, supporting our broader evaluation based on a statistically determined sample of posts which achieved 93% agreement between human reviewers and the AI model.
The final analysis involved a statistically validated number of posts reviewed by the same annotator.
Testing concluded that the model performed exceptionally well, matching or even exceeding human reviewer consistency across most categories.
Accuracy (Percentage of correctly classified instances): 94.7%.
Precision (Percentage of correctly assigned true label counts by GPT): 79.5%.
Recall (Percentage of instances classified as true by humans and also classified as true by GPT): 86.1%.
F1 Score (A single percentage combining accuracy and recall, with higher values indicating better classification): 82.6%.
The model’s performance was evaluated by an internal statistical analyst, concluding its results were strong, benchmarked against similar academic studies.
Despite the model’s impressive performance, misclassifications in the analysis are inevitable.
We believe the classification process employing OpenAI’s API is thorough, transparent, defensible, and bolsters rigorous journalism.
Quick Guide
Please contact us about this story
show
The best public interest journalism relies on direct accounts from people of knowledge.
If you have anything to share about this subject, please contact us confidentially using the following methods:
Secure Messaging in Guardian App
The Guardian app includes a tool for sending story tips. Messages are end-to-end encrypted and concealed within routine operations on the Guardian mobile app, preventing any observer from realizing the communication.
If you haven’t yet downloaded the Guardian app, do so here (iOS/Android) and navigate to ‘Secure Messaging’ in the menu.
SecureDrop, Instant Messengers, Email, Phone, and Mail
If you can safely access the TOR network without being detected, you can send messages and documents to the Guardian through our <a href=\"https://www.theguardian.com/securedrop\">SecureDrop platform</a>.
Finally, our guide at <a href=\"https://www.theguardian.com/tips\">theguardian.com/tips</a> details various secure contact methods, outlining the pros and cons for each.
Over the weekend, hackers gained access to the X account of the beloved character Elmo and used it to post offensive racist and anti-Semitic threats, along with profane remarks about Jeffrey Epstein. On Monday, Sesame Workshop was actively working to reclaim full control of its Red Character account.
“Unknown hackers have compromised Elmo’s X account, posting derogatory messages, including anti-Semitic and racist content. We are striving to regain complete control of our account,” a spokesman for Sesame Workshop stated on Monday. Sesame Workshop is the nonprofit organization that produces Sesame Street and Elmo.
The account was breached over the weekend, and instead of its usual positive posts, Elmo’s 650,000 followers were confronted with anti-Semitic threats and racist commentary. The account also included profane references to convicted sex trafficker Epstein, revealing additional details about him to the US government, amidst the ongoing debate over the FBI’s findings that Epstein committed suicide. Although these tweets were swiftly deleted, Elmo’s account still hosts a link to a Telegram channel that celebrates the hacking.
X, owned by billionaire Elon Musk, has not yet responded to requests for comment.
Recently, Elmo’s social media accounts have focused on mental health awareness. Last year, the eternally three-and-a-half-year-old red monster prompted responses from Joe Biden and Chance the Rapper with a simple check-in: “Elmo is just checking in! What is everyone doing?”
X has faced ongoing challenges in safeguarding high-profile accounts, which can cause significant harm with a single tweet. Last year, various British MPs and international entities experienced similar breaches, with their accounts being used to mock cryptocurrency.
ICE can provide a method for long-term message storage in cold climates
Anton Petrus/Getty Images
Messages can be preserved in ice for millennia simply by altering the shape and positioning of the internal bubbles present.
Mengjie Song and his team at China’s Beijing University of Technology were exploring ice formation when they discovered that it influences the size and shape of bubbles encapsulated within. For instance, they found that freezing a layer of water between plastic sheets resulted in either oval or needle-shaped foam, depending on the freezing rate.
The researchers assigned specific bubble sizes, shapes, and positions to represent characters in Morse and binary code. By controlling the freezing rate of water, they created ice that conveyed messages through the embedded bubbles.
Transforming this ice image to grayscale revealed that white areas indicated the presence of bubbles, while black areas indicated the absence of them. This allowed computers to identify the size and location of bubbles for message decoding.
Currently, the amount of information that can be stored in conventional ice cubes is minimal with existing technology. However, Song suggests that by manipulating foam in materials like plastic, greater capacities could be achieved.
He mentions that beyond the novelty of reading messages embedded in ice cubes used in beverages, this research has diverse potential applications. “The strength of this study lies in its capability to sustain information over extended periods in frigid environments like the Arctic and Antarctic,” Song notes.
He envisions a future where bubbles could safely contain ozone for food preservation or act as carriers for slow-release medications. He is particularly intrigued by how bubbles can inhibit ice formation on airplane wings and how this knowledge could inform behavior in lunar environments.
However, Qiang Tang from the University of Sydney is more skeptical about the practical implications of this research, asserting that significant information can be long-term stored on hard drives and paper.
He comments, “This represents a novel method for conveying messages, but from a security and encryption standpoint, I see little value in it—unless, of course, a polar bear has something to communicate,” he adds.
Are your messages still marked as read? A recent study suggests that not getting replies to your messages could be linked to your texting habits. Many people use abbreviations in their text messages, but research from Stanford University suggests that this may signal a lack of interest in the conversation. Read more here.
The study also found that individuals who use abbreviations in their texts are less likely to receive responses, and the use of abbreviations can disrupt the flow of conversations.
According to the study authors at the American Psychological Association, “Perceived lack of effort can negatively impact interpersonal interactions when texting.” Lead researcher David Fang, a PhD student at Stanford University, emphasized the importance of appearing sincere and building social connections, especially in the early stages of a relationship or when trying to make a good impression.
This study, published in the Journal of Experimental Psychology: General, analyzed conversations from 37 countries involving over 5,300 text messagers. Researchers examined how message recipients reacted in conversations with and without abbreviations on dating apps and chat rooms.
Contrary to popular belief, using abbreviations does not make individuals seem relaxed; rather, it can make them appear dishonest. The research revealed that 80% of participants believed others would not mind their use of abbreviations, while 4% incorrectly thought abbreviations would have a positive impact.
“We initially thought abbreviations were perceived as casual and familiar, but we were surprised to find they actually created negative perceptions of the sender,” explained Huang, one of the study’s researchers.
In addition to abbreviations, using emojis to convey messages can also lead to misunderstandings. A recent study found that the intended meaning of emojis can be easily misinterpreted. Learn more here.
Prior research has shown that using humble punctuation marks at the end of texts can make readers perceive the message as less authentic than a message without punctuation. Readers considered exclamation points as the most honest way to convey excitement in texts.
Polarized light can erase messages encoded in quantum holograms
Hong Liang, Wai Chun Wong, Tailing Ang, Jensen Lee 2024
The quantum evanescence phenomenon makes it possible to embed secure messages in holograms and selectively erase parts of them even after they have been transmitted.
Quantum optical signals are inherently secure information carriers: any interception of the message destroys the fragile quantum states that encode it. To harness this without the use of bulky devices, Jensen Lee Researchers from the University of Exeter in the UK MetasurfaceIt is a 2D material engineered with special properties to create quantum holograms.
Holograms encode complex information that can be restored when light is shone on it. For example, when light hits a 2D holographic paper card at the right angle, a 3D image appears. To create quantum holograms, researchers encoded information in the quantum state of particles of light, or photons.
First, they used a laser to emit two photons from a special crystal that were tightly bound by quantum entanglement. The photons traveled along separate paths, with only one encountering the metasurface along the way. Thousands of tiny components on the metasurface, like nano-sized bumps, altered the photon’s quantum state in a preprogrammed way, encoding a holographic image into it.
The partner photon encountered a polarizing filter, which controlled which parts of the hologram appeared and which disappeared. The first photon’s state was a superposition of holograms, so it contained different variations of the message at the same time. Because the photons were in an entangled state, polarizing the second photon affected the image the other photon created when it hit the camera. For example, a test hologram contained the letters H, D, V, and A, but adding a filter for horizontal polarization caused the letter H to disappear from the final image.
Li says metasurfaces could be used to encode more complex information into photons, for example as part of quantum cryptography protocols. He calls the research SPIE Optics + Photonics Conference August 21st, San Diego, California.
“Everyone dreams of quantum technology going from square metres on a table to being compact enough to fit in a smartphone, and metasurfaces seem like a good way to achieve that. [about that]” Andrew Forbes A researcher at the University of Witwatersrand in South Africa, said quantum holograms like the one used in this experiment could also be used to image tiny biological structures in the rapidly expanding medical field.
SNine years ago today, Joaquin Oliver was murdered in the hallway outside his Florida classroom. He was one of 17 students and staff killed in America's deadliest high school shooting. On Wednesday, lawmakers in Washington, D.C., will hear his voice recreated by artificial intelligence on the phone, asking them why they haven't done more about the gun violence epidemic.
“It's been six years and you haven't done anything. You can't stop the shootings that have happened since then,” he said of the Valentine's Day 2018 shooting at Marjory Stoneman Douglas High School in Parkland. A message from Oliver, who was 17 at the time of his tragic death, reads:
“I came back today because my parents used AI to recreate my voice and call you. Other victims like me have also received countless calls demanding action. How many calls will it take to care? How many dead voices will I hear before I finally hear it?”
Oliver is one of six people who lost their lives to firearms, and his voice is about to be heard again. He's issuing a call to action in an innovative online gun reform campaign launched today. shot line.
Parkland victim Joaquin Oliver
“How many dead voices will we hear before we finally hear it?”
Sorry, your browser does not support audio. However, you can download and listen here $https://uploads.guim.co.uk/2024/02/13/TheShotline_AI_JoaquinOliver_Call_to_Congress.mp3
A project by two activist groups formed in the wake of the Parkland shooting and creative communications agency MullenLowe, it leverages AI technology to generate direct messages from shooting victims themselves.
The voices are “trained” using deep machine learning from audio clips provided by family members. The resulting recordings are ready to go directly to the people in Congress who have the power to take action against gun violence. Website visitors enter their zip code and choose the message they want to send to their elected representatives.
“We all hear children's voices in our heads. Why don't lawmakers need to hear them too?” said Mike Song, whose 15-year-old son Ethan died in an accident involving a missing gun.
Ethan's message, like Oliver's, is straightforward. “Children like me die every day. It's time to act. It's time to pass laws that protect children from unsafe guns. At the end of the day, it's about helping people. It’s your job to pass responsible gun control, or we’ll find someone to do it.”
Other voices recreated for the Shotline project include that of 10-year-old Ujiyah Garcia, a victim of the 2022 Uvalde Elementary School shooting in Texas. Akira DaSilva, 23, was killed in the 2018 Waffle House shooting in Tennessee. Jaycee Webster, 20, was shot and killed by an intruder in his Maryland home in 2017. And in 2014, Mike Bohan committed suicide with a gun he could buy in 15 minutes.
Vaughn's death, who suffered from depression, sparked a movement that led to passage of Maryland's first Red Flag gun control.
Six years after Oliver's murder, it is by design that Oliver's voice is at the forefront of the campaign. One of his two groups behind this effort is march for our livean activist group formed by Stoneman Douglas students that sparked global protests after Parkland.
The Shotline campaign uses AI to generate audio messages from gun violence victims. Photo: shot line
the other one is, Change references, was founded by the teenager's parents, Manny and Patricia Oliver. They have been relentlessly advocating for gun reform since his son was murdered.
“We wanted this to be a powerful message,” Patricia Oliver said. “Joaquin has his own energy, his own image, and that's what keeps him alive. I'm so proud of Joaquin, he's the driving force that drives us forward.”
She admits the process of recreating her son's voice for 56 seconds was mentally taxing. The Olivers searched their phones and computers for videos containing Joaquin's statements and asked her sister Andrea, other relatives and girlfriend Tori to do the same.
“It was difficult to make out his exact voice because of the noise in the background,” she said. “In one video, he was in the pool and we were talking and the sound of the water was distracting.”
Eventually, we assembled enough clips for our engineers to work with, and after a long period of fine-tuning, we received the final “draft.”
“When I played it, it was incredibly shocking and a lot of different emotions came up. We had been listening to videos of Joaquín talking about the past, and now he's in a situation where he is today, very emotional. We talk about recent things,” she said.
“I know this is just a fantasy and not the truth. But in that moment, you forget what you're listening to, why you're listening, and he just says, 'Hello, Mom, how are you?' I just hope from the bottom of my heart that you just say, “?”. once again. “
Ethan's mother, Christine Song, said she felt the same painful emotions when she heard her son “talk” again six years after his death.
“It brings you back to that day, the last words your child said to you before leaving your life,” she said.
“Honestly, I just sat there and sobbed, because I knew he would never come back. But the Olivers, and my husband, and people like us all have one thing in common: What we're saying is that we go out every day and fight for respect for our children, and we're actually fighting for your children and grandchildren.”
The Songs are pressuring federal lawmakers to pass the Connecticut bill. ethan's lawrequires safe storage of firearms in the home.
“We have promised that we will not stop until we can create a cultural shift in this country where gun owners make safe storage of their weapons second nature,” said Kristen Song. Ta. “You might think that's enough because the coffins of our dead children are piling up, but when it comes to Republicans in Congress, they just don't listen.”
To create voice and calls, MullenLowe talking baby For E*Trade's Super Bowl commercial, we partnered with AI specialist Edisen, with teams in the US and Sweden working on the project.
Snippets of audio “trained” on speech patterns and tonality were fed through Eleven Labs’ generative voice AI platform, and the reconstructed voices generated voice calls from text-to-speech scripts.
“There's a lot of talk about AI right now, but this is a beautiful example of what AI can actually achieve, and a very human achievement,” says Mirko, AI creative designer at Stockholm-based Edisen.・Mr. Lempert said.
“This project was very moving and showed me how different our world is, because in my country we are not exposed to it.” [gun violence] That's the situation. That was a wake-up call. ”
Last week, the Federal Communications Commission banned robocalls using AI-generated voices after Joe Biden's voice was imitated in a fake phone call to voters in New Hampshire.
MullenLowe said Shotline calls are exempt because they are not auto-dialed, are made to a landline and are provided with a callback number.
In the hills south of Rome is Italy’s premier nuclear physics laboratory, the Frascati National Laboratory. It has all the equipment you’d expect from a state-of-the-art scientific facility, including giant magnets, powerful particle accelerators, and exposed electrical wires strung throughout. Many of the researchers here are trying to unlock the secrets of the Standard Model, the best theory of how reality works at the most fundamental level. And then there’s the room where Catalina Cruceanu is keeping watch over a small box of lentils.
Admittedly, this is not at all normal behavior for a physicist, but Cruceanu explains why the equipment and methods of nuclear physics cause lentils and other organisms to constantly emit extremely weak photons and particles. We hope to solve the 100-year-old mystery. light’s. Some people think that these “biophotons” are not important. Others argue that they are a subtle form of lentil communication. Cruceanu leans towards the latter position, and even has a hunch that the pulses between pulses may contain secret quantum signals. “These are just the first steps, but it looks like it’s going to be very interesting,” she says.
There are already hints that living things exploit quantum phenomena, and there is also inconclusive evidence that quantum phenomena have features in things like photosynthesis and the way birds move. But lentils may be the most surprising example of quantum biology yet, because their complex behavior is poorly understood, he says. Michal Shifra At the Czech Academy of Sciences in Prague. “That would be great,” Shifra says. “If that’s true.” Because so many living things emit biophotons, such a discovery could indicate that quantum effects are ubiquitous…
This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.
Strictly Necessary Cookies
Strictly Necessary Cookie should be enabled at all times so that we can save your preferences for cookie settings.
