MultiSearch Tag Explorer MultiSearch Tag Explorer - Explore tags and search results by aéPiot - aéPiot: Independent SEMANTIC Web 4.0 Infrastructure (Est. 2009). High-density Functional Semantic Connectivity with 100/100 Trust...

#DELVE
semantic-search.headlines-world.com/advanced-sea...
#LẠC #SƠN #HÒA #BÌNH
semantic-search.aepiot.com/advanced-sea...
ĐIỆN #BIÊN
multi-search-tag-explorer.allgraph.ro/advanced-sea...
aepiot.ro

#Delve did the security compliance on #LiteLLM, an #AI project hit by #malware

techcrunch.com/2026/03/25/delve-did-the...

#cybersecurity

Anyhow, working on AI stuff is always fun because it's like: "Look at the little dude I made, walking around and doing stuff!!"

But who's this mysterious NPC who'll show up in delve 0.6.0? :o

#delve #roguelikedev

Original post on techcabal.com

Why being polite to AI might be hurting your results The researchers tested how different tones, ranging from very polite to very rude, affect ChatGPT-4o's performance on multiple-choice questi...

#Delve #into #AI #Africa #AI #adoption #Artifical #Intelligence […]

[Original post on techcabal.com]

The irony writes itself...
#Delve #LiteLLM #Cybersecurity #SOC2 #SecurityCompliance

I've spent the last few hours discussing this point with several LLMs.

philosophics.blog/2026/03/24/100-human-con...

If the AI tells are so well-known and predictable, why don't we simply reverse them?

tl;dr? […]

[Original post on mastodon.social]

a joke and a tragedy, lol 😂

#ai #llm #delve #security

#Delve accused of misleading customers with ‘fake #compliance’

techcrunch.com/2026/03/22/delve-accused...

#privacy #HIPAA #GDPR

Delve faces serious allegations of providing fake compliance evidence, potentially exposing clients to legal risks. Transparency in tech compliance is more crucial than ever. #TechNews #Compliance #Delve Link: thedailytechfeed.com/delve-faces-...

A design dilemma I am having with #delve right now is that bows have unlimited ammo, but I've also added special arrows.

I gave unlimited ammo because I didn't want a player who wants to play an archer to have to micromanage their inventory.

#roguelikedev

Delve faces allegations of providing fake compliance evidence, potentially exposing clients to legal risks. Transparency in compliance processes is crucial. #Delve #Compliance #RegulatoryRisk Link: thedailytechfeed.com/delve-faces-...

Amerika, 22 yaşındaki Selin'i konuşuyor! - ekovizyon.com.tr Dünyanın en prestijli teknoloji girişimlerinden birini kuran Selin Kocalar, Forbes’un 30 Altında ki 30 isim listesine girerek büyük ses getirmişti. Ancak şimdi Delve üzerinde başlatılan soruşturma gün...

Amerika, 22 yaşındaki Selin'i konuşuyor!

Delve üzerinde başlatılan #soruşturma gündemi sarsıyor.

www.ekovizyon.com.tr/amerika-22-y...

#amerika #selinkocalar #girişimcilik #delve

Amerika, 22 yaşındaki Selin'i konuşuyor!

Delve üzerinde başlatılan #soruşturma gündemi sarsıyor.

www.ekovizyon.com.tr/amerika-22-y...

#amerika #selinkocalar #girişimcilik #delve

An acquaintance telephoned me to debate whether it’s possible to hire a cut-rate SOC2 vendor without yourself being a shitbird, but our call was interrupted when they received an email from their #Delve rep advising them of some new breach or leak or something. I love this industry.

I don't know, man. Forging any documents is a bad look, but these ones were literally about security and data privacy. The fact that when you were asked, can you be trusted?, you responded by frantically splashing together some crude bootlegs is, like, too on the nose to our time. It's because of you unrealistic-growth-at-any-cost yo-yos that our industry has become a chaotic, stressful, and nihilistic place. Everyone's now being told to cut corners, break things, and ship ever flimsier merchandise in pursuit of ever improbable lucre, and it's grinding good people down, and I hate it, very much. I'm not saying that crippling your businesses with years of litigation or throwing all of you dingdongs in jail will actually help anything, but let me have this, as a treat.

Well, you got me going about #Delve. www.linkedin.com/posts/cowper...

I don't know, man. Forging any documents is a bad look, but these ones were literally about security and data privacy. The fact that when you were asked, can you be trusted?, you responded by frantically splashing together some crude forgeries is, like, too on the nose to our time. It's because of you unrealistic-growth-at-any-cost yo-yos that our industry has become a chaotic, stressful, and nihilistic place. Everyone's now being told to cut corners, break things, and ship ever flimsier merchandise in pursuit of ever improbable lucre, and it's grinding good people down, and I hate it, very much. I'm not saying that crippling your businesses with years of litigation or throwing all of you dingdongs in jail will actually help anything, but let me have this, as a treat.

Well, you got me going about #Delve. www.linkedin.com/posts/cowperthwait_soc2-...

Item description of an artifact in Delve. It reads +2 Tempest. Tempest: a longsword forged long ago to battle Greater Old Belesorad when they assailed these lands. Personally wielded in battle by Aenab Ruler of SPring. THere are many who say this weapon is of ill omen and brings with it a curse. This is a sword. It deals 1d8 slashing, 1d8 holy damage. It is rustproof.

I'd honestly forgotten that I'd implemented (a basic version of) randomized artifact weapons in Delve

#delve #roguelikedev

Delve - Fake Compliance as a Service - Part I How Delve managed to falsely convince hundreds of customers they were compliant and then lied about it when exposed and called out

#Delve claims ‘Compliance in days’. But #DeepDelver #AICPNAY tells very different story, claiming “@getdelve built a machine designed to make clients complicit w/o their knowledge, to manufacture plausible deniability while producing exactly the opposite”. substack.com/home/post/p-...

Delve Faked 494 Compliance Reports. The EU AI Act Was Designed to Prevent Exactly This. A leaked spreadsheet revealed that GRC platform Delve generated 494 near-identical SOC 2 reports with pre-written auditor conclusions. The EU AI Act's conformity assessment framework was designed to p...

A leaked Google Sheet just exposed how Delve, a Y Combinator-backed GRC platform with $32M in Series A funding, generated 494 near-identical SOC 2 audit reports.

They used the same boilerplate, and the same "independent" conclusions

systima.ai/blog/delve-c... #delve #ai

This evening I am adding magic arrows that have special effects.

#delve #roguelikedev

Mirage Delve farming blueprint for Path of Exile: routes, scarab combos, fossil flips, and a repeatable profit loop to boost currency. Read: haplogamingchef.blogspot.com/2026/03/path...

#PathOfExile #PoE #Delve #Mirage #Fossils #Scarabs #CurrencyFarming #GGG #ActionRPG #Gaming

Another version of the lunge animation 🤔 #delve #roguelikedev

Right? I still remember your top notch design of the lava for #Delve ! Golden, I'm telling ya!

As African economies digitise rapidly, cybercrime is evolving just as quickly. Malware that once took skilled programmers weeks or months to build can now be generated in minutes using AI-powered coding tools, enabling cybercriminals to launch cheaper, faster, and large-scale attacks, often targeting businesses and consumers coming online for the first time. The shift is captured in the HP Wolf Security Threat Insights Report, released by the security unit of technology manufacturer HP Inc, which shows attackers shifting from carefully engineered exploits toward a strategy built on speed and volume. By combining AI-assisted coding with modular malware kits, often purchased cheaply on underground forums, cybercriminals can now generate thousands of slightly different malware samples and launch them across the internet within minutes. Rather than investing time in building technically perfect malware, attackers are increasingly relying on large numbers of ‘good enough’ attacks that are inexpensive, automated, and difficult to detect individually. In some cases identified by HP researchers, hackers purchase ready-made malware components for less than $10 and use automated tools to modify them repeatedly. Even if most of these attacks fail, the sheer scale means that a small number of successful infections can still produce significant financial returns. The implications are particularly significant for emerging digital economies. Across Africa, businesses are rapidly adopting cloud services, digital payments, and AI-driven infrastructure. But that rapid digital adoption also expands the region’s cyber-attack surface. According to the HP report, organisations across the continent experience an average of 3,153 cyberattacks weekly—about 60% higher than the global average—suggesting that attackers are actively targeting environments where cybersecurity practices are still maturing. For small and medium-sized enterprises (SMEs), the economic imbalance behind these automated attacks is especially stark. While cybercriminals can assemble malware campaigns for only a few dollars, the damage from a single successful breach can be devastating. Cybercrime is estimated to cost African economies roughly $10 billion annually, and for smaller businesses, the consequences can be existential. In South Africa, for example, a study shows that around 22% of SMEs hit by ransomware attacks ultimately shut down. In this new era of automated cybercrime, the low cost of launching attacks contrasts sharply with the potentially catastrophic cost of defending against them. ## **The shift from precision to scale** For many years, the most dangerous cyberattacks were often the most technically sophisticated ones. Highly skilled hackers would craft malware capable of quietly infiltrating networks, stealing sensitive data, or spreading across systems undetected. These attacks required time, expertise, and careful testing. Cybercriminals are adopting a software-like approach to attacks, using automated coding tools to generate, test, and deploy new malware variants within minutes. This speed-over-perfection strategy allows them to launch hundreds or thousands of slightly different attacks, increasing the chance some will bypass defenses. In one HP-identified case, attackers hid malware inside a Scalable Vector Graphic (SVG) image—a file type made of lines and shapes rather than pixels—which browsers open automatically and email filters often trust, letting the malicious code slip past initial security checks. In Nigeria, the average organization now faces roughly 4,701 cyberattacks weekly. Most of these are not highly sophisticated, hand-crafted hacks but automated scripts designed to scan systems and exploit a single weak point. ## **AI-assisted coding accelerates malware development** AI-assisted coding tools—often described as “vibe coding”—are becoming a major driver of change in cybercrime. These tools can generate working software code from simple prompts, helping developers build applications faster. But the same capability is now being exploited by cybercriminals to create malicious programs with far less effort than before. In the past, writing malware required advanced technical skills and weeks or months of work to design programs that could infiltrate systems and evade antivirus detection. AI tools have lowered that barrier dramatically. Attackers can now generate key malware components, such as “loaders”—small programs that enter a victim’s computer and download additional malicious software—in just seconds. Even when the AI-generated code is imperfect, attackers can quickly modify it or produce many variations until one works. Each version appears slightly different to security systems, making it harder for traditional antivirus tools that rely on known malware signatures to detect them. This constant variation acts like a digital disguise, allowing some attacks to slip through defenses—something reflected in HP’s findings that 14% of email threats in late 2025 bypassed at least one email security scanner before being stopped. ## **The rise of modular “flat-pack” malware** Another trend highlighted in the HP report is the rise of modular malware kits, sometimes called “flat-pack malware.” Instead of building malicious software entirely from scratch, attackers now assemble it from pre-built components available online. These modules can include loaders, credential-stealing tools, ransomware functions, and command-and-control systems. By combining different pieces, cybercriminals can quickly create customised malware packages for specific campaigns. Automated coding tools make this even easier by generating scripts that connect the modules or help disguise them from security systems. This modular approach lowers the technical barrier to launching cyberattacks. People with limited programming knowledge can assemble working malware using components purchased or downloaded from underground forums. As a result, the number of potential attackers is growing rapidly, making the cybersecurity landscape more complex and unpredictable. ## **Brand mimicry and the rise of digital “evil twins”** While automated coding helps attackers build malware faster, they still rely heavily on deception to persuade victims to install it. One of the most effective techniques highlighted in the HP report involves brand mimicry. Cybercriminals are becoming increasingly adept at creating fake websites that closely resemble legitimate platforms used by millions of people. Services such as Microsoft Teams, Zoom, and Booking.com are common targets because users trust them and frequently download their software. Attackers replicate these sites with remarkable precision. Logos, colors, layouts, and even the wording used on official pages are copied to create convincing “evil twin” versions of the real websites. In the Microsoft Teams “Piggyback” campaign (2025–2026), hackers used SEO poisoning to manipulate search results so that anyone searching for “download Microsoft Teams” was directed to a fake website that looked identical to the official page. When users clicked “Download,” they received a fully functional copy of Teams—but it was secretly bundled with a hidden malware file called OysterLoader, giving attackers access while leaving the main app working as expected. Similarly, the Booking.com “ClickFix” and “I Paid Twice” campaigns in November 2025 relied on psychological trickery targeting hotel staff and travelers. Emails mimicked legitimate guest complaints, directing staff to a fake Booking.com portal claiming their browser was malfunctioning. Following the prompt to “fix” the issue—a tactic known as ClickFix—installed malware such as PureRAT or XWorm, giving attackers covert access to their systems. In Africa, banks are often the main targets of brand-mimicry attacks because they provide direct access to money. In one example known as the “Help Desk” scam in Nigeria and South Africa, criminals create fake social media accounts using the logos and branding of major banks such as United Bank for Africa, Standard Bank, and First Bank of Nigeria. When customers complain online about failed transactions, the fake accounts quickly respond and direct them to a cloned banking website designed to steal their login details. Cybercriminals boost the reach of these fraudulent sites using search-engine poisoning, exploiting algorithm weaknesses to push malicious pages to the top of search results. A user searching for a popular software installer may unknowingly click on one of these fake sites, believing it to be legitimate. Once the victim downloads the installer from the counterfeit page, the attack begins. In many cases, the real software will install and function normally, reinforcing the illusion that the download was legitimate. However, a hidden malicious program may also be installed in the background. One example is a loader known as OysterLoader, which acts as a backdoor into the infected system. While the user continues using the legitimate application, attackers gain remote access to the computer. The rise of AI-assisted malware demonstrates that modern cyberattacks rely as much on deception as on technical sophistication. As these methods continue to spread worldwide, the takeaway is clear: effective cybersecurity needs to go beyond simply detecting threats and instead adopt proactive strategies designed to anticipate and counteract deception at every stage.

Cybercrime costs Africa $10 billion a year. AI is about to make that number bigger. The HP Wolf Security Threat Insights Report, released by the security unit of technology manufacturer HP Inc, sho...

#Delve #into #AI #Artifical #Intelligence #cybercrime #Hackers

Origin | Interest | Match

Someone reported a game crash in Delve D: How am I supposed to go to sleep now!

#delve #roguelikedev

Artificial intelligence (AI) could add as much as $136 billion in productivity gains across Africa, according to Microsoft. Unlocking that value, however, depends on whether countries can enable the secure, seamless flow of data across their borders. “We only have about 1%or 2% of global compute power in Africa,” Akua Gyekye, Government Affairs Director for Microsoft Africa, told TechCabal in an interview on Wednesday.“If we’re serious about diffusing AI across the continent, we need infrastructure. But we also need the right policy environment.” AI systems demand enormous computing power, yet Africa controls less than 1% of global capacity. With limited access to hyperscale infrastructure, researchers, startups, and governments struggle to deploy advanced AI tools at scale. In response, African governments have repeatedly called on global cloud providers such as Microsoft to build local data centres. As of mid-2025, Africa has 223 data centres across 38 countries, less than 0.02% of the world’s total of more than 11,800. Microsoft was the first massive-scale cloud operator to launch local data centres in Africa (2019). These are organised into two primary regions: South Africa North (Johannesburg) and South Africa West (Cape Town). But Gyekye argues that a purely national approach may not be economically viable or strategically sound. “A regional approach often makes more sense,” she said. “If you build a data centre in Nigeria, how can the rest of West Africa benefit? If you build in South Africa, how do neighbouring countries plug in?” Microsoft’s existing South African cloud regions already serve customers across Southern Africa, raising a broader policy question on whether other African countries allow their citizens’ data to reside in neighbouring jurisdictions while still enforcing their own data-protection laws. That is where the conversation shifts from infrastructure to sovereignty. ## **Data sovereignty vs. data silos** Over the past decade, Africa has seen a rapid surge in data-protection regulation. In 2012, only 12 countries had data-protection laws; at least 46 have now enacted legislation or regulatory frameworks, and more than 40 now operate dedicated Data Protection Authorities. The African Union’s Convention on Cyber Security and Personal Data Protection, which came into force in 2023, has helped speed up alignment efforts across regional blocs. Enforcement has also intensified. Regulators in countries such as Nigeria, South Africa and Kenya are issuing record fines and, in some cases, criminal penalties for violations. In July 2024, the Federal Competition and Consumer Protection Commission (FCCPC), in a landmark joint action with the Nigeria Data Protection Commission, fined Meta $220 million for “intrusive” data practices, unauthorised data sharing, and abusing its market dominance. In late 2023 and again in 2025, Kenya’s Office of the Data Protection Commissioner (ODPC) fined several digital lenders between KES 2 million ($15,500) and KES 5 million ($38,760) for “shaming” borrowers by contacting their phone contacts without permission. Yet Gyekye warned that well-intentioned sovereignty policies can create data silos. “Some governments feel data is safest when it is physically within their borders,” she said. “But cybersecurity threats don’t respect borders. In many cases, cloud providers can offer higher levels of protection than local hosting.” The issue, she argued, is not about weakening sovereignty but about modernising it. If a country’s laws can travel with its data, enforced through contractual, technical and regulatory safeguards, then physical location becomes less critical. In Ethiopia, strict localisation requirements mandate that certain data be stored on local servers. While such measures aim to protect citizens, they also complicate regional expansion for banks, fintech startups and e-commerce firms seeking to scale across East Africa. Safaricom’s expansion into Ethiopia illustrates how differing national data rules complicate the “super-app” model. To launch M-Pesa, the company had to comply with Ethiopia’s strict local data-processing requirements, different from the provisions of Kenya’s Data Protection Act of 2019. “Data silos don’t just affect Microsoft,” Gyekye noted. “They impact regional banks, small and medium enterprises, and startups that aim to expand beyond a single market.” The African Continental Free Trade Area (AfCFTA) is designed to promote cross-border collaboration, with the African Union estimating it could boost intra-African trade by over 50%. Yet, digital trade, payments, logistics, e-commerce, and AI-driven services all depend on the seamless movement of data across borders. “Without cross-border data collaboration, the promise of AfCFTA cannot be fully realised,” Gyekye said. She argued that harmonisation does not require identical laws across all 54 countries. Instead, interoperability and mutual recognition frameworks could allow data to flow securely between jurisdictions with comparable standards. “There isn’t a one-size-fits-all model,” she said. “Countries can maintain their sovereignty and self-determination. But there must be pathways for data to move.” ## **The sovereign cloud model** To bridge the gap between sovereignty and scale, Microsoft is advocating a “sovereign cloud” model, which allows data to be hosted outside a country’s physical borders while remaining subject to that country’s laws and regulatory controls. In practice, this often involves encryption safeguards, local key management, strict access controls, and contractual guarantees ensuring that only authorised entities can access sensitive information. For example, a major South African tier-one bank such as Standard Bank or FirstRand may run large-scale data workloads on Microsoft Azure while relying on Thales CipherTrust to manage encryption keys locally. “The question becomes: who holds the keys?” Gyekye explains. “Not Microsoft. The customer or the government.” The approach reflects a broader industry trend toward modular cloud architectures. Sensitive personal data may remain locally stored, while anonymised or aggregated data can be processed regionally or globally for analytics and AI training. Amazon Web Services (AWS) uses a modular approach called AWS Outposts, which brings cloud hardware into the customer’s data centre. In January 2026, AWS announced the expansion of its 2nd-generation Outposts racks to Nigeria, Morocco, and Senegal. For policymakers, the model offers a compromise: maintain legal oversight while accessing larger pools of compute power. Still, Gyekye insists that infrastructure alone will not unlock Africa’s AI dividend. Microsoft’s strategy on the continent also focuses heavily on skills development and connectivity. Through its Airband initiative, a global program aimed at expanding internet access to underserved and rural communities, the tech giant says it has connected more than 117 million Africans to broadband, surpassing its initial targets. It is also investing in AI skilling programmes for students, developers and policymakers. “Technology is just a tool,” she said. “If people don’t know how to use it, or don’t trust it, the economic benefits won’t materialise.” Language accessibility remains a significant barrier to AI adoption. On February 4, 2026, Microsoft unveiled expanded AI capabilities in 39 African languages through Paza—Swahili for “to project” or “to raise your voice.” The initiative aims to close the “AI divide” by ensuring that AI tools are locally relevant and usable. “If AI doesn’t speak your language, you won’t use it,” Gyekye noted. ## **Big Tech and the African agency** Concerns remain that the deep involvement of global technology firms could weaken African digital sovereignty. Much of this worry stems from the power imbalance between trillion-dollar multinational companies and developing digital economies. The U.S. CLOUD Act allows American authorities to compel U.S.-based tech companies to hand over data they control, even if that data is stored abroad. In practice, this means a Nigerian citizen’s information held in a Lagos data centre operated by a U.S. firm could, in theory, be accessed by Washington without the Nigerian government’s approval. Gyekye acknowledges the concern but emphasises multi-stakeholder governance. When Kenya launched its national AI strategy in March 2025 at the Kenyatta International Convention Centre in Nairobi, she said, it invited local startups, researchers, civil society groups and international companies, including Microsoft, to the decision-making process where these groups helped define the six key pillars of the strategy. “You can’t regulate what you don’t understand,” she said. “Governments must lead, but they need input from industry, academia and civil society.” In her view, it is African governments, rather than foreign firms, that must set the direction because they have the deepest understanding of how these decisions affect their people. “It has to be government-led,” she said. “Only governments know the specific priorities of their citizens.” ## **The $136 billion question** If Africa can align infrastructure, governance and skills, Microsoft’s estimate of $136 billion in productivity gains may prove conservative. AI applications in agriculture, healthcare, logistics and education could dramatically improve efficiency and inclusion. Imagine satellite data helping farmers predict crop yields across East Africa. Or AI-powered diagnostics supporting overstretched rural clinics. Or cross-border fintech platforms scaling seamlessly under AfCFTA rules. But if restrictive data localisation rules proliferate without interoperability mechanisms, the continent risks fragmentation. The result could be duplicated infrastructure, higher costs and slower innovation. For Gyekye, success would mean removing unnecessary data silos while respecting national sovereignty. “Data can flow, keeping sovereignty top of mind,” she said. “That’s how we build digital economies that benefit everyone, from small businesses to multinationals, from policymakers to the man on the street.”

Microsoft says AI could add $136 billion to Africa if data can move freely Artificial intelligence (AI) could add as much as $136 billion in productivity gains across Africa, according to Microsoft...

#Delve #into #AI #Africa #AI #Africa #Microsoft

Origin | Interest | Match

Screenshot of ascii roguelike game. The player is fighting a gas spore (an 'e' on screen) and a centipede (a 'c' on screen)

I heard you like roguelikes with fewer bugs and grammar mistakes so I've released delve 0.5.4.

Also: emergency doors

#delve #roguelikedev

Screenshot of a scene in an ASCII roguelike game where the player character is fighting several blobs, one of which is grappling them. The grappler is shown in a slightly different colour.

A little quality of life fix: when you are grappled, you can only melee attack the monster grappling you, but when you are surrounded by several monsters of the same type it was hard to tell which one was grappling you, so now I'm slightly altering its colour:

#delve

Smart monsters will now seek out teleport traps when they are frightened and trying to flee from danger.

(this was pretty simple -- just a matter of changing the weight for teleport traps for pathfinding)

#delve #roguelikedev

Art made in A4 for Immortals RPG, Lost Legacies Gaming

Enter a world of sword and sorcery where adventurers try their luck in a grim world superintended by immortals and their followers.

#ImmortalsRPG #ODnD #WhiteBox #delve #dungeon #journey #ttrpg #fantasy #osr #oldschool #dnd #crawlstilho