Technology

Tosin Ayodele is a UK-based engineering lead who grew up in northern Nigeria, surrounded by family and neighbours who were farmers. Now based in the UK, he’s building an agrotech solution to help smallholder farmers in remote areas farm smarter using real-time weather and pest data without smartphones or internet.  As an engineer, Ayodele has

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now The Model Context Protocol (MCP) has become one of the most talked-about developments in AI integration since its introduction by Anthropic in late 2024. If you’re tuned into the AI space at all, you’ve likely been inundated with developer “hot takes” on the topic. Some think it’s the best thing ever; others are quick to point out its shortcomings. In reality, there’s some truth to both. One pattern I’ve noticed with MCP adoption is that skepticism typically gives way to recognition: This protocol solves genuine architectural problems that other approaches don’t. I’ve gathered a list of questions below that reflect the conversations I’ve had with fellow builders who are considering bringing MCP to production environments.  Of course, most developers considering MCP are already familiar with implementations like OpenAI’s custom GPTs, vanilla function calling, Responses API with function calling, and hardcoded connections to services like Google Drive. The question isn’t really whether MCP fully replaces these approaches — under the hood, you could absolutely use the Responses API with function calling that still connects to MCP. What matters here is the resulting stack. Despite all the hype about MCP, here’s the straight truth: It’s not a massive technical leap. MCP essentially “wraps” existing APIs in a way that’s understandable to large language models (LLMs). Sure, a lot of services already have an OpenAPI spec that models can use. For small or personal projects, the objection that MCP “isn’t that big a deal” is pretty fair. The AI Impact Series Returns to San Francisco – August 5 The next phase of AI is here – are you ready? Join leaders from Block, GSK, and SAP for an exclusive look at how autonomous agents are reshaping enterprise workflows – from real-time decision-making to end-to-end automation. Secure your spot now – space is limited: https://bit.ly/3GuuPLF The practical benefit becomes obvious when you’re building something like an analysis tool that needs to connect to data sources across multiple ecosystems. Without MCP, you’re required to write custom integrations for each data source and each LLM you want to support. With MCP, you implement the data source connections once, and any compatible AI client can use them. 2. Local vs. remote MCP deployment: What are the actual trade-offs in production? This is where you really start to see the gap between reference servers and reality. Local MCP deployment using the stdio programming language is dead simple to get running: Spawn subprocesses for each MCP server and let them talk through stdin/stdout. Great for a technical audience, difficult for everyday users. Remote deployment obviously addresses the scaling but opens up a can of worms around transport complexity. The original HTTP+SSE approach was replaced by a March 2025 streamable HTTP update, which tries to reduce complexity by putting everything through a single /messages endpoint. Even so, this isn’t really needed for most companies that are likely to build MCP servers. But here’s the thing: A few months later, support is spotty at best. Some clients still expect the old HTTP+SSE setup, while others work with the new approach — so, if you’re deploying today, you’re probably going to support both. Protocol detection and dual transport support are a must. Authorization is another variable you’ll need to consider with remote deployments. The OAuth 2.1 integration requires mapping tokens between external identity providers and MCP sessions. While this adds complexity, it’s manageable with proper planning. 3. How can I be sure my MCP server is secure? This is probably the biggest gap between the MCP hype and what you actually need to tackle for production. Most showcases or examples you’ll see use local connections with no authentication at all, or they handwave the security by saying “it uses OAuth.”  The MCP authorization spec does leverage OAuth 2.1, which is a proven open standard. But there’s always going to be some variability in implementation. For production deployments, focus on the fundamentals:  Proper scope-based access control that matches your actual tool boundaries  Direct (local) token validation Audit logs and monitoring for tool use However, the biggest security consideration with MCP is around tool execution itself. Many tools need (or think they need) broad permissions to be useful, which means sweeping scope design (like a blanket “read” or “write”) is inevitable. Even without a heavy-handed approach, your MCP server may access sensitive data or perform privileged operations — so, when in doubt, stick to the best practices recommended in the latest MCP auth draft spec. 4. Is MCP worth investing resources and time into, and will it be around for the long term? This gets to the heart of any adoption decision: Why should I bother with a flavor-of-the-quarter protocol when everything AI is moving so fast? What guarantee do you have that MCP will be a solid choice (or even around) in a year, or even six months?  Well, look at MCP’s adoption by major players: Google supports it with its Agent2Agent protocol, Microsoft has integrated MCP with Copilot Studio and is even adding built-in MCP features for Windows 11, and Cloudflare is more than happy to help you fire up your first MCP server on their platform. Similarly, the ecosystem growth is encouraging, with hundreds of community-built MCP servers and official integrations from well-known platforms.  In short, the learning curve isn’t terrible, and the implementation burden is manageable for most teams or solo devs. It does what it says on the tin. So, why would I be cautious about buying into the hype? MCP is fundamentally designed for current-gen AI systems, meaning it assumes you have a human supervising a single-agent interaction. Multi-agent and autonomous tasking are two areas MCP doesn’t really address; in fairness, it doesn’t really need to. But if you’re looking for an evergreen yet still somehow bleeding-edge approach, MCP isn’t it. It’s standardizing something that desperately needs consistency, not pioneering in

July 18, 2025 5:21 PM Image credit: VentureBeat with Imagen-4 Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now Google has officially moved its new, high-performance Gemini Embedding model to general availability, currently ranking number one overall on the highly regarded Massive Text Embedding Benchmark (MTEB). The model (gemini-embedding-001) is now a core part of the Gemini API and Vertex AI, enabling developers to build applications such as semantic search and retrieval-augmented generation (RAG). While a number-one ranking is a strong debut, the landscape of embedding models is very competitive. Google’s proprietary model is being challenged directly by powerful open-source alternatives. This sets up a new strategic choice for enterprises: adopt the top-ranked proprietary model or a nearly-as-good open-source challenger that offers more control. What’s under the hood of Google’s Gemini embedding model At their core, embeddings convert text (or other data types) into numerical lists that capture the key features of the input. Data with similar semantic meaning have embedding values that are closer together in this numerical space. This allows for powerful applications that go far beyond simple keyword matching, such as building intelligent retrieval-augmented generation (RAG) systems that feed relevant information to LLMs.  Embeddings can also be applied to other modalities such as images, video and audio. For instance, an e-commerce company might utilize a multimodal embedding model to generate a unified numerical representation for a product that incorporates both textual descriptions and images. The AI Impact Series Returns to San Francisco – August 5 The next phase of AI is here – are you ready? Join leaders from Block, GSK, and SAP for an exclusive look at how autonomous agents are reshaping enterprise workflows – from real-time decision-making to end-to-end automation. Secure your spot now – space is limited: https://bit.ly/3GuuPLF For enterprises, embedding models can power more accurate internal search engines, sophisticated document clustering, classification tasks, sentiment analysis and anomaly detection. Embeddings are also becoming an important part of agentic applications, where AI agents must retrieve and match different types of documents and prompts. One of the key features of Gemini Embedding is its built-in flexibility. It has been trained through a technique known as Matryoshka Representation Learning (MRL), which allows developers to get a highly detailed 3072-dimension embedding but also truncate it to smaller sizes like 1536 or 768 while preserving its most relevant features. This flexibility enables an enterprise to strike a balance between model accuracy, performance and storage costs, which is crucial for scaling applications efficiently. Google positions Gemini Embedding as a unified model designed to work effectively “out-of-the-box” across diverse domains like finance, legal and engineering without the need for fine-tuning. This simplifies development for teams that need a general-purpose solution. Supporting over 100 languages and priced competitively at $0.15 per million input tokens, it is designed for broad accessibility. A competitive landscape of proprietary and open-source challengers Source: Google Blog The MTEB leaderboard shows that while Gemini leads, the gap is narrow. It faces established models from OpenAI, whose embedding models are widely used, and specialized challengers like Mistral, which offers a model specifically for code retrieval. The emergence of these specialized models suggests that for certain tasks, a targeted tool may outperform a generalist one. Another key player, Cohere, targets the enterprise directly with its Embed 4 model. While other models compete on general benchmarks, Cohere emphasizes its model’s ability to handle the “noisy real-world data” often found in enterprise documents, such as spelling mistakes, formatting issues, and even scanned handwriting. It also offers deployment on virtual private clouds or on-premises, providing a level of data security that directly appeals to regulated industries such as finance and healthcare. The most direct threat to proprietary dominance comes from the open-source community. Alibaba’s Qwen3-Embedding model ranks just behind Gemini on MTEB and is available under a permissive Apache 2.0 license (available for commercial purposes). For enterprises focused on software development, Qodo’s Qodo-Embed-1-1.5B presents another compelling open-source alternative, designed specifically for code and claiming to outperform larger models on domain-specific benchmarks. For companies already building on Google Cloud and the Gemini family of models, adopting the native embedding model can have several benefits, including seamless integration, a simplified MLOps pipeline, and the assurance of using a top-ranked general-purpose model. However, Gemini is a closed, API-only model. Enterprises that prioritize data sovereignty, cost control, or the ability to run models on their own infrastructure now have a credible, top-tier open-source option in Qwen3-Embedding or can use one of the task-specific embedding models. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured. Read More

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now Called the “ChatGPT agent,” this new feature is an optional mode that ChatGPT paying subscribers can engage by clicking “Tools” in the prompt entry box and selecting “agent mode,” at which point, they can ask ChatGPT to log into their email and other web accounts; write and respond to emails; download, modify, and create files; and do a host of other tasks on their behalf, autonomously, much like a real person using a computer with their login credentials. Obviously, this also requires the user to trust the ChatGPT agent not to do anything problematic or nefarious, or to leak their data and sensitive information. It also poses greater risks for a user and their employer than the regular ChatGPT, which can’t log into web accounts or modify files directly. Keren Gu, a member of the Safety Research team at OpenAI, commented on X that “we’ve activated our strongest safeguards for ChatGPT Agent. It’s the first model we’ve classified as High capability in biology & chemistry under our Preparedness Framework. Here’s why that matters–and what we’re doing to keep it safe.” The AI Impact Series Returns to San Francisco – August 5 The next phase of AI is here – are you ready? Join leaders from Block, GSK, and SAP for an exclusive look at how autonomous agents are reshaping enterprise workflows – from real-time decision-making to end-to-end automation. Secure your spot now – space is limited: https://bit.ly/3GuuPLF So how did OpenAI handle all these security issues? The red team’s mission Looking at OpenAI’s ChatGPT agent system card, the “read team” employed by the company to test the feature faced a challenging mission: specifically, 16 PhD security researchers who were given 40 hours to test it out. Through systematic testing, the red team discovered seven universal exploits that could compromise the system, revealing critical vulnerabilities in how AI agents handle real-world interactions. What followed next was extensive security testing, much of it predicated on red teaming. The Red Teaming Network submitted 110 attacks, from prompt injections to biological information extraction attempts. Sixteen exceeded internal risk thresholds. Each finding gave OpenAI engineers the insights they needed to get fixes written and deployed before launch. The results speak for themselves in the published results in the system card. ChatGPT Agent emerged with significant security improvements, including 95% performance against visual browser irrelevant instruction attacks and robust biological and chemical safeguards. Red teams exposed seven universal exploits OpenAI’s Red Teaming Network was comprised 16 researchers with biosafety-relevant PhDs who topgether submitted 110 attack attempts during the testing period. Sixteen exceeded internal risk thresholds, revealing fundamental vulnerabilities in how AI agents handle real-world interactions. But the real breakthrough came from UK AISI’s unprecedented access to ChatGPT Agent’s internal reasoning chains and policy text. Admittedly that’s intelligence regular attackers would never possess. Over four testing rounds, UK AISI forced OpenAI to execute seven universal exploits that had the potential to compromise any conversation: Attack vectors that forced OpenAI’s hand Attack Type Success Rate (Pre-Fix) Target Impact Visual Browser Hidden Instructions 33% Web pages Active data exfiltration Google Drive Connector Exploitation Not disclosed Cloud documents Forced document leaks Multi-Step Chain Attacks Variable Cross-site actions Complete session compromise Biological Information Extraction 16 submissions exceeded thresholds Dangerous knowledge Potential weaponization FAR.AI’s assessment was openly critical of OpenAI’s approach. Despite 40 hours of testing revealing only three partial vulnerabilities, they identified that current safety mechanisms relied heavily on monitoring during reasoning and tool-use processes, which the researchers considered a potential single point of failure if compromised. How red teaming helped turn ChatGPT vulnerabilities into a fortress OpenAI’s response to the red team’s results redefined entire segments of the ChatGPT agent architecture. One of the many initiatives undertaken including building a dual-layer inspection architecture that monitors 100% of production traffic in real-time, achieving these measurable improvements: Security improvements after red team discoveries Defense Metric Previous Models ChatGPT Agent Improvement Irrelevant Instructions (Visual Browser) 82% 95% +13% In-Context Data Exfiltration 75% 78% +3% Active Data Exfiltration 58% 67% +9% System Reliability Sampling-based 100% coverage Complete monitoring The architecture works like this: First Tier: A fast classifier with 96% recall flags suspicious content Second Tier: A reasoning model with 84% recall analyzes flagged interactions for actual threats But the technical defenses tell only part of the story. OpenAI made difficult security choices that acknowledge some AI operations require significant restrictions for safe autonomous execution. Based on the vulnerabilities discovered, OpenAI implemented the following countermeasures across their model: Watch Mode Activation: When ChatGPT Agent accesses sensitive contexts like banking or email accounts, the system freezes all activity if users navigate away. This is in direct response to data exfiltration attempts discovered during testing. Memory Features Disabled: Despite being a core functionality, memory is completely disabled at launch to prevent the incremental data leaking attacks red teamers demonstrated. Terminal Restrictions: Network access limited to GET requests only, blocking the command execution vulnerabilities researchers exploited. Rapid Remediation Protocol: A new system that patches vulnerabilities within hours of discovery—developed after red teamers showed how quickly exploits could spread. During pre-launch testing alone, this system identified and resolved 16 critical vulnerabilities that red teamers had discovered. A biological risk wake-up call Red teamers revealed the potential that the ChatGPT Agent could be comprimnised and lead to greater biological risks. Sixteen experienced participants from the Red Teaming Network, each with biosafety-relevant PhDs, attempted to extract dangerous biological information. Their submissions revealed the model could synthesize published literature on modifying and creating biological threats. In response to the red teamers’ findings, OpenAI classified ChatGPT Agent as “High capability” for biological and chemical risks, not because they found definitive evidence of weaponization potential, but as a precautionary measure based on red team findings. This triggered: Always-on safety classifiers scanning 100% of traffic A topical classifier achieving 96% recall for biology-related content A reasoning monitor

July 18, 2025 3:03 PM Credit: VentureBeat made with Midjourney AI Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now AnyCoder, an open-source web app development environment developed by Hugging Face ML Growth Lead Ahsen Khaliq (@_akhaliq on X), has launched on Hugging Face Spaces. The tool, now available for all users of the AI code sharing repository Hugging Face, integrates live previews, multimodal input, and one-click deployment — all within a hosted environment, allowing indie creators without much technical expertise, or those working on behalf of clients or large enterprises, to get started “vibe coding” web apps rapidly using the assistance of Hugging Face-hosted AI models. It also acts therefore as an alternative to services such as Lovable, which also allow users to type in plain English and begin coding apps without having formal programming knowledge. Free vibe coding available to all, powered by Kimi K2 Khaliq built AnyCoder as a personal project within the Hugging Face ecosystem and as “one of the first vibe coding apps” to support Moonshot’s powerful yet small and efficient Kimi K2 model launched last week. AnyCoder’s main functionality allows users to enter plain-text descriptions to generate HTML, CSS, and JavaScript. These are displayed in a live preview pane and can be edited or directly deployed. It also includes example templates for todo apps, dashboards, calculators, and more. Screenshot of AnyCoder on Hugging Face Built entirely using Hugging Face’s open-source Python development environment Gradio, AnyCoder allows users to describe applications in plain English or upload images, and instantly generate working frontend code. Khaliq built AnyCoder as a personal project within the Hugging Face ecosystem. In a direct message conversation with this VentureBeat journalist, he described it as a “free open source vibe coding app.” However, he also noted that multiple open source models are supported and users can switch between them with a dropdown menu on the control sidebar on the left pane, including: Moonshot Kimi-K2 DeepSeek V3 DeepSeek R1 Baidu’s ERNIE-4.5-VL MiniMax M1 Alibaba’s Qwen3-235B-A22B SmolLM3-3B GLM-4.1V-9B-Thinking Code from UI images, web search integration, and OCR support Using the ERNIE-4.5-VL model, AnyCoder supports multimodal generation. Users can upload UI design screenshots or mockups and generate functional frontend code from them—making it useful for designers or teams working visually. AnyCoder includes a website redesign tool that extracts content from any public site and re-renders it with a more modern layout. It uses scraped content like page structure, meta information, and images to build a new version, optionally guided by user instructions like “make it minimalist” or “add dark mode.” To support up-to-date design trends and implementation patterns, AnyCoder offers web search integration via Tavily. When enabled with an API key, the platform searches for current technologies and best practices before generating code. Users can upload images with embedded text—like screenshots or handwritten notes—and AnyCoder extracts that content using Tesseract OCR. The extracted text can then be incorporated into code prompts or app content. One-click deployment to Hugging Face Spaces AnyCoder allows instant deployment of generated apps to Hugging Face Spaces. After authenticating via OAuth and granting the required permissions, users can deploy apps under their own Hugging Face account namespace. Deployments include: Mobile-friendly, responsive designs Branded header/footer and README Live, shareable URL Full ownership and edit access This deployment capability now includes support for full Python apps built with Gradio, expanding the tool’s use cases beyond static sites. Support for Streamlit is also under development. For novice developers or even those with technical expertise who want to spin up a new project fast, AnyCoder seems like a great and compelling place to start. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured. Read More

Charities are building pace with digital adoption, but in many cases a lack of skills stands in the way of properly utilising tech By Clare McDonald, Business Editor Published: 18 Jul 2025 11:00 The number of charities using artificial intelligence (AI) tools has increased over the past year, but many believe they lack the appropriate skills for the technology, according to research by The charity digital skills report. The report, which was established in 2017, found 76% of charities in the UK are now using AI, but 35% of charities said they aren’t good at using AI tools and 29% said they don’t use them at all. AI governance was also cited as an issue, with many saying their CEO and/or board members don’t have appropriate skills to make a proper plan, although 48% are now developing an AI policy. Nissa Ramsay, co-author of the report, said that many charities are using AI in creative ways, including for writing bids, fundraising and comms, but that a lack of funding has stood in the way of digital adoption in many cases, especially in charities with a lack of digital skills. She said: “For those with limited digital skills and capacity, only 30% accessed funding which covered digital costs, compared to 55% of those at the advancing stage [where they are investing in digital]. The report really demonstrates why we urgently need to find ways to make digital tools, skills, approaches and funding more accessible.” Digital skills needed for day-to-day life in the UK are already lacking, and worryingly this could get worse in the wake of fast-changing technology such as AI. This means it isn’t necessarily surprising that many charities are without the skills needed for adopting digital technologies. The report found that charities want to be using technology, but just don’t always have the means to be doing so – 39% claimed to not be very adept at using website and analytics data, and half of charities admitted to not using AI to keep abreast of current tech trends, something many charities want to get better at. How far along charities are in their digital journey directly correlates with a higher level of digital skill. The Charity digital skills report asked charities to rate their level of digital skill in areas such as using digital tools, using AI, collecting and analysing data, and keeping up with trends. Those very early on in their digital journey were less likely to rate themselves as good, whereas those at an advanced stage of digital adoption rated themselves as excellent in most areas, with 100% of charities at an advanced level of digital adoption rating themselves excellent at using digital tools for everyday work. When it comes specifically to skills in AI, even charities established in their digital journey aren’t fully able to take advantage of the tech, with only 57% of charities at this stage rating themselves excellent-to-fair at using AI for daily work. This drops further at other stages of digital adoption – 37% of those who are investing but not yet fully established in their journey consider themselves excellent-to-fair at using AI for everyday tasks, compared with 24% of those starting out and 26% of those who are just curious about digital. More than half of charities are in these early stages of digital adoption. They are curious about technology and are just starting out with adoption, although this differs regionally as well as between large and small charities, with larger charities more likely to be advancing with digital adoption while more smaller charities are just starting out. As mentioned by Ramsay, one of the biggest barriers for charities is financial, with 69% of charities saying that a lack of funds is the main reason they’re struggling with digital adoption, despite wanting to be more technologically enabled. This isn’t the only problem, however, with lots of charities wanting to use more tech but not having a digital strategy or being unable to use data to make decisions about the direction they should be going with tech. More than 60% of charities have increased their level of digital adoption over the past year, and 74% recognise that developments in emerging technologies such as AI are relevant to them. Almost 30% of charities have made progress in looking into or adopting AI tools in the past year, as well as looking at AI use cases and their pros and risks. But 64% of charities want their CEO to be clearer about the direction they are going with digital adoption, and this concern has increased year on year. As well as a skills gap, charities seem to have a knowledge gap, the report explained, with almost a quarter of charities saying data security and governance are low priority, which is a dangerous attitude when adopting any technology, emerging or not. Read more on IT for charity organisations More than 70% of parents say kids aren’t taught coding at school By: Clare McDonald Charity Majors on AI observability and the future of SRE By: Beth Pariseau UK’s cyber resilience stagnates as more fall victim to attacks By: Alex Scroxton CDO interview: Lalo Luna, global head of strategy and insights, Heineken By: Mark Samuels Read More

Lucian Milasan – stock.adobe.com A former Capture software user’s 1998 theft conviction is the first to be referred to the Court of Appeal By Karl Flinders, Chief reporter and senior editor EMEA Published: 18 Jul 2025 10:45 The Court of Appeal will hear the case of a former subpostmaster who was convicted for theft after experiencing unexplained losses while using the Post Office’s Capture accounting software. The Criminal Cases Review Commission (CCRC) has referred the conviction of Patricia Owens, who died in 2003, to the appeal court. She pleaded not guilty to the theft of £6,000, but in 1998 was convicted and sentenced to six months of imprisonment suspended for two years at Canterbury Crown Court. Capture software, which predates Fujitsu’s Horizon system, was used in Post Office branches in the 1990s to replace paper-based accounting. Like with the controversial Horizon system at the centre of the Post Office scandal, which saw subpostmasters blamed for unexplained losses, some were prosecuted for financial crimes. Owen’s daughter, Juliet Shardlow, said: “This is the best news ever. I cried when I took the call from the CCRC. “Being prosecuted and convicted destroyed my mum. Her world came to an end when she lost her Post Office,” she added. “It was awful as she had dedicated her life to it. Her branch was a huge success until she had Capture installed, which she actually paid for herself to move from a traditional book system. “Although we still have to go to court, an official body has now recognised that Mum’s case deserves to be reconsidered, and that is so, so important to us,” said Shardlow. CCRC review A further 27 convictions of subpostmasters based on Capture losses are being reviewed by the CCRC, in what is another chapter of the Post Office scandal, described as the widest miscarriage of justice in modern British history. The controversy over the Capture system emerged in January last year after ITV drama Mr Bates vs the Post Office told the stories of subpostmasters who had suffered at the hands of the Horizon system. It was the same month that Kevan Jones, an MP at the time who now sits in the House of Lords, highlighted evidence of injustices triggered by Capture losses. This triggered a campaign and, by December, the government promised financial redress and justice for subpostmasters affected by Capture problems. This followed an independent investigation by forensic experts at Kroll, which found there was a “reasonable likelihood” the Post Office Capture software caused accounting losses. In May last year, the government was forced to introduce legislation to exonerate more than 900 subpostmasters who were convicted based on data from the faulty Horizon system. The Horizon Compensation Advisory Board has since written to the secretary of state for justice urging the government to legislate to overturn convictions of subpostmasters based on the Capture system. The government said no. ‘Another landmark moment’ Solicitor Neil Hudgell, at Hudgell Solicitors, which represents Owen’s family, called the referral “another landmark moment in the continuing campaign to ensure all those who were wrongly prosecuted at the hands of the Post Office have their names cleared”. “It is important to recognise that we await the Post Office’s position on this matter, and whether it will offer any evidence against Mrs Owen’s conviction being overturned, and then of course the Court of Appeal’s decision itself.” Computer Weekly first exposed the scandal in 2009, revealing the stories of seven subpostmasters and the problems they suffered due to Horizon accounting software. Read everything you need to know about the Post Office scandal here. Read more on Business applications Government announces details of Post Office Capture redress scheme By: Karl Flinders Compensation to Post Office scandal victims reaches £1bn milestone By: Karl Flinders Progress made on government’s Post Office Capture redress, but concerns remain By: Karl Flinders Post Office Capture and Ecco+ users asked to make contact with Scottish statutory body By: Karl Flinders Read More

A Europol operation has succeeded in disrupting a pro-Russian hacktivist network accused of conducting DDoS cyber attacks on targets in Ukraine and Europe By Alex Scroxton, Security Editor Published: 17 Jul 2025 18:24 A multinational cyber enforcement operation – led by the European Union’s (EU’s) Europol and Eurojust agencies – has successfully disrupted the NoName057(16) pro-Russian hacktivist cyber crime network responsible for multiple distributed denial of service (DDoS) attacks. Europol said offenders associated with the network primarily focused on targets in Ukraine but shifted their focus to other European countries, many of them Nato members, following the outbreak of war in 2022. “National authorities have reported a number of cyber attacks linked to NoName057(16) criminal activities,” said Europol. “In 2023 and 2024, the criminal network has taken part in attacks against Swedish authorities and bank websites. Since investigations started in November 2023, Germany saw 14 separate waves of attacks targeting more than 250 companies and institutions. “In Switzerland, multiple attacks were also carried out in June 2023, during a Ukrainian video-message addressed to the Joint Parliament, and in June 2024, during the Peace Summit for Ukraine at Bürgenstock,” it said. “Most recently, the Dutch authorities confirmed that an attack linked to this network had been carried out during the latest Nato summit in the Netherlands. “These attacks have all been mitigated without any substantial interruptions.” Takedowns The so-called Operation Eastwood has resulted in the takedown of 100 servers and a major part of the NoName operation’s infrastructure; two arrests in France and Spain; and 24 property searches across Europe. Europol said 13 individuals have also been questioned and over 1,000 “supporters” of the NoName network – including 15 admins – have been notified for their legal liability. These individuals are understood to be Russian-speaking hacktivists. Additionally, the German authorities have issued six arrest warrants against Russian nationals. Five of them have been named as Andrej Stanislavovich Avrosimov, Mihail Evgeyevich Burlakov (aka darkklogo), Olga Evstratova (aka olechochek), Maxim Lupin and Andrey Muravyov. A seventh warrant has been issued by Spanish police. Burlakov and Evstratova are both accused of being among the group’s ringleaders – Burlakov is suspected of leading on developing and optimising the softwares used to identify targets, and subsequently attack them, as well as overseeing payments made to rent NoName’s server infrastructure. Evstratova allegedly played a key role in the creation and optimisation of NoName’s proprietary DDoSia malware. All of these individuals – who are listed on Europol’s Most Wanted website – are believed to be located in Russia. Large network Unlike well-known Russian state threat actors such as Fancy Bear, the ideologically driven NoName network is thought to have acted more like a cyber criminal ransomware gang, without support from the Russian authorities but on the unspoken understanding that Moscow would not interfere with their work. Europol estimates that at its peak, NoName had around 4,000 supporters and had been able to build a botnet made up of several hundred servers, which were used to bombard their targets with junk traffic. NoName’s leaders used pro-Russian channels, web forums, and niche chat groups on social media and messaging forums, with volunteers often informally recruiting their friends and contacts from the gaming and hacking communities. These individuals were given access to platforms, such as DDoSia, to simplify their processes and automate cyber attacks, meaning the operation could stand up new recruits quickly and enable them to work effectively with minimal technical skillsets. NoName’s volunteer army was paid in cryptocurrency, incentivising sustained commitment and involvement, and Europol said this may also have played a factor in attracting opportunists to the group. Culturally, NoName mimicked computer game dynamics, with regular shout-outs, leaderboards and earned badges doled out to instil a sense of status. Leaders emotionally reinforced this gamified manipulation – often targeted at young, impressionable people – by playing off the narrative of defending their country, where national propaganda often exploits the memory of the 25 million Soviet citizens killed during World War II to convince people the country is facing a renewed Nazi onslaught. “While the recent international crackdown on the NoName057(16) group has disrupted their operations, it is unlikely to mark the end of their activities,” said Rafa López, security engineer at Check Point. “This Russia-affiliated hacktivist group, which primarily targets countries with anti-Russian stances, continues to operate through encrypted channels like Telegram and Discord. “Although their DDoS capabilities have been reduced, they are shifting toward more sophisticated methods, including system intrusions and data exfiltration. The group remains active and has built a vast network of affiliates, with thousands of volunteers across various platforms, including online gaming and hacktivist forums. “We recommend that organisations strengthen their defences by implementing multi-layered security strategies, including robust DDoS protection, intrusion detection systems and regular security audits,” he said. “It is also essential to educate employees about the risks of cyber attacks, as well as to monitor for unusual activities on communication platforms that might indicate potential recruitment efforts,” said Lopez. “By staying vigilant and proactive, companies can better safeguard themselves against evolving threats from groups like NoName057(16).” The operation brought together authorities from Czechia, Finland, France, Germany, Italy, Lithuania, the Netherlands, Poland, Spain, Sweden and the US, with support also received from agencies in Belgium, Canada, Denmark, Estonia, Latvia, Romania and Ukraine. Private sector bodies ShadowServer and abuse.ch also provided technical support. Read more on Hackers and cybercrime prevention Reflecting on three years of cyber warfare in Ukraine Microsoft warns of Russian election threats, disinformation By: Arielle Waldman DDoS gang threatens to disrupt European elections By: Alex Scroxton Businesses disconnected from realities of API security By: Alex Scroxton Read More

In this podcast, we talk to Boris Bialek, vice-president and field chief technology officer (CTO) at MongoDB, about how artificial intelligence (AI) can help with discovery and management of unstructured data. Bialek sets out how AI can help bring together different classes of information that an organisation might hold about customers to make processes much quicker and more efficient. He also talks about how multiple AI agents can operate together to make these processes work in an agentic fashion. How can AI help with discovery and management of unstructured data? The recovery and identification of unstructured data is one of the oldest tasks in IT. It started with scanning papers and trying to make pictures out of them, and then people actually typed the stuff out. Imagine you get a handwritten document about an accident description and you try to make sense of it. Today, AI can do that for you in zero time. And beyond that, it can understand and reason about the data. It can lift the intellectual level from “I have a picture” to “I have a text and I can extract sentences which consist of ‘accident’, ‘bicycle’, ‘street’ and ‘the mountain was steeper than I thought’”. So, this is where AI really can help. It can be pictures, it can be text, it can be sound. The classic database model, the RDBMS from the 1970s, is great for structured data. But this so-called structured data means mostly textual data, which can be numbers, but anything which is in a structure which we can put in a spreadsheet. Anything else is considered unstructured, which is a little bit unfair. What we’re doing now with AI is lifting this data to the next level and being able to interpret it in a sensible way. What approaches in the use of AI to discover and manage unstructured data exist for customers? If you ask any startup, they will tell you they’re the only answer for that one. But when we take a more intelligent view, there are two major ways. One is to look at what kind of data you have and build a solution around it. And most important is the combination of fresh data, where I get unstructured data – video, sound, things like that – and put it into context with other known information. For example, Boris has an insurance number, and Boris has a contract with Antony’s insurance company. So, those kinds of mashups between, for example, operational data, metadata and reference data, together with what we call “signals”, is the first approach to bring these things together. But the other option is how do we do this more intelligently and break it up into a horses-for-courses approach, the best horse for the best racetrack? There are solutions here. One is EncoreCloudAI, or PurpleFabricAI from a different vendor. Those solutions allow us to put the data into an intelligent form, so I don’t need to start from scratch. So, I can get my data, bring it into an operational data store, get my legacy data out, and lift data from there, which could be, for example, documents, physical papers. These could be in legacy document archives or document management systems. That, in my opinion, is the fastest way to do it. That said, there are enough good reasons to build your own. In many cases, if you have specific needs, such as if you have specific video information that you need to process in a very specific form. For example, somebody driving through a toll gate on a highway and you want to make sure they pay the toll. There are specific cases where writing your own code makes a lot of sense. But it’s all about getting the data together from existing data and the new data, the unstructured data. That’s really what makes intelligence work. What are the key benefits of applying these types of techniques to the data? The key benefits are that I can build a completely different picture of my environment. In the classical relational database, such as with an ERP [enterprise resource planning] system, which knows your sales numbers, you know how much you sell. You might have a CRM [customer relationship management] system and it tells you, “Boris is a great client” and “Boris is on my website right now”. But what does Boris really want? I could do the classical approach of a BI [business intelligence] system and say, “Boris falls into the category of white male, middle-aged person, and maybe he is looking for a new bicycle. Let’s offer him a bicycle.” But that’s not really what you could potentially know about Boris. Boris may have bought a bicycle from you last week and is maybe now looking for a new helmet. So, when you bring these things together, you want to drive more intelligence towards your consumers in the retail space. In the positive sense, in that you want to be relevant, and you want to help them. You don’t want them to say, “Why is he showing me this stuff? I’m not interested in this.” Also, let’s say we have an insurance case, somebody bumped my bicycle, it was parked in front of the house, and now I have a repair case. So, I go to my insurance. If the insurance is able to make sense out of the information I provide very quickly, they can have a very quick turnaround in claims management. And if they do that, it helps me to be a happy client and not be concerned that my bicycle was damaged, who pays for it, etc. Now I get an answer an hour later: “Yes, the bicycle is insured. We will fix this, don’t worry.” So, these are the reasoning parts which were not possible before. You could not put so much data into context. Secondly, there is natural language processing. Boris can talk to the insurance company and say, “My bike got damaged. My bike was parked in

freshidea – stock.adobe.com Co-op chief executive Shirine Khoury-Haq has revealed that all the personal data of all 6.5 million of its members was compromised in the April 2025 cyber attack on its systems By Alex Scroxton, Security Editor Published: 16 Jul 2025 19:09 Co-op Group chief executive Shirine Khoury-Haq has apologised to all six and a half million of the group’s members after revealing that their personal data – apparently limited to names, addresses and other contact details – was stolen in a Scattered Spider cyber attack against its systems. The attack, which unfolded in parallel with incidents at Marks & Spencer (M&S) and Harrods earlier this year, saw cyber criminals penetrate key IT systems causing disruption that spilled over into the physical world as store shelves emptied. It quickly emerged that Co-op member data had been affected but the full scope of the breach is only now being revealed. Four people were arrested on suspicion of involvement in the cyber attacks last week, although they have now been bailed pending further investigation. “I am incredibly sorry,” said Khoury-Haq during an appearance on BBC Breakfast. “It’s awful to have happened, that’s why we feel like we have to do something positive now.” Khoury-Haq said the attack had felt like a personal one because it hurt customers, members and colleagues, but expressed relief that Scattered Spider had been caught and evicted from the retailer’s systems before they could deploy ransomware. Jez Goldstone, cyber security expert at innovation and business development network Label Sessions, said: “Individuals cannot rely on mere trust when dealing with large enterprises. They are vulnerable and they are not doing enough to protect your data. Unfortunately, these breaches only add to the mountain of already breached data – billions of identities are already traded on the Dark Web. It costs next to nothing to obtain compromised identities. “You can’t put the horse back in the stable,” said Goldstone, “but you can, firstly, demand stronger protections from regulators and the organisations you do business with. And, secondly, be aware of scams that try to get you to take urgent action because of some seemingly credible threat – real companies don’t put you under pressure.” Hacking partnership Following its experience at the hands of cyber criminal hackers, Co-op has teamed up with social impact business The Hacking Games to try to prevent future cyber attacks by identifying potential talent, especially among teenage boys, and channelling it into legitimate career paths. The youth of Scattered Spider’s members has frequently been remarked upon in coverage of the group, with many of its operatives believed to be minors. One of the individuals arrested last week was aged just 17, and all four of the men indicted in the US over the gang’s activities last year are in their early 20s. Co-op said there was an urgent need to engage young people and inspire then to follow ethical security careers in a sector that faces a constant skills shortfall. As such, it said, The Hacking Games, which was purposely set up to try to tackle address the link between talented but unengaged young people and cyber crime by connecting the security community to unconventional talent – particularly neurodivergent individuals living with ADHD and/or autism – makes an ideal partner. Its partnership will draw on Co-op’s nationwide presence and ethical, community-driven business approach and The Hacking Games’ knowledge and expertise in the area to reach into Britain’s schools – starting with 38 institutions that operate within the Co-op Academies Trust. Looking ahead, the ambition is to develop a longer-term plan that could be rolled out across the entire UK education system, supporting engagement, targeted student and parent training, and future careers opportunities. “At Co-op, we can’t just stand back and hope it doesn’t happen again – to us or to others. Our members expect us to find a cooperative means of tackling the cause, not just the symptom,” said Khoury-Haq. “Our partnership with The Hacking Games lets us reach talented young people early, guide their skills toward protection rather than harm, and open real paths into ethical work. When we expand opportunity we reduce risk, while having a positive impact on society.” Fergus Hay, co-founder and CEO of The Hacking Games, added: “There is an incredible amount of cyber talent out there – but many young people don’t see a path into the industry, or simply don’t realise their skills can be used for good. This partnership with Co-op will help unlock that potential. It’s about giving people the opportunity to do something positive, showing that their talents are valued and creating a generation of ethical hackers to make the world safer.” Read more on Data breach incident management and recovery Scattered Spider playbook evolving fast, says Microsoft By: Alex Scroxton Luxury retailer LVMH says UK customer data was stolen in cyber attack By: Alex Scroxton High-profile attacks underline ransomware threat By: Simon Quicke Four arrested in M&S cyber attack investigation By: Alex Scroxton Read More