Category: Biz Tech

China’s AI sector is accelerating beyond reliance on imported semiconductors, leaning into algorithmic ingenuity and open-source collaboration to offset hardware constraints and project growing soft power abroad.Chinese firms face significant barriers in securing the most advanced GPUs, due to U. S. export restrictions. Rather than compete on sheer processing capacity, companies such as DeepSeek have turned to architectural innovation. Its R1 large language model, released under an openly shared licence, delivers performance comparable to Western counterparts such as GPT-4 and

Anthropic has agreed to pay US $1.5 billion to settle a class-action lawsuit filed by several authors who alleged the company illegally downloaded and stored pirated copies of their books to train its AI chatbot Claude. A federal court filing has been submitted to U. S. District Judge William Alsup in San Francisco, who is scheduled to review the agreement. If approved, the payout will likely represent the largest copyright recovery in U. S. history. Payment to authors is estimated

OpenAI is launching a platform to connect workers fluent in artificial intelligence with employers, using AI to match skills and opportunities OpenAI builds AI-Powered Jobs Hub, positioning itself against LinkedIn. Set to debut by mid-2026, the service aims to streamline recruitment by ensuring candidates’ AI competencies align closely with employer requirements.In a blog post, Fidji Simo, OpenAI’s chief executive of Applications, said the OpenAI Jobs Platform, as it is known, will feature tracks tailored for both large firms and local

Hon Hai Precision Industry Co. reported August sales of NT$606.5 billion, marking a 10.6 per cent increase year-on-year. The Taiwanese manufacturer, a key production partner for Nvidia’s AI-enabled servers, anticipates both sequential and year-on-year growth in third-quarter revenue, underpinned by a sharp rise in cloud and networking product shipments. Hon Hai’s August sales climb 10.6 per cent amid AI demand reflects continued strength in AI infrastructure spending.Growth in August follows a slower expansion in July, when sales rose only

A Chinese state-sponsored cyber-espionage group known as Salt Typhoon almost surely infiltrated Canadian telecommunications systems during the campaign that exfiltrated data belonging to millions of Americans. The intrusion, part of the broader Chinese-backed hackers almost certainly breached Canadian networks operation, raises alarm over the growing reach and sophistication of cross-border cyber-threats.The FBI and Canada’s Centre for Cyber Security have attributed the breach in Canada to Salt Typhoon, which covertly embedded itself into the infrastructure of at least one telecom provider.

Dream Security and other cybersecurity analysts have identified a widely distributed, Iran-linked spear-phishing campaign that exploited a compromised Ministry of Foreign Affairs mailbox in Paris. The operation impersonated legitimate Omani diplomatic communication to deliver malware-laden Word documents to government, diplomatic, and multilateral institutions across Europe, Africa, Asia, the Americas, and international organisations such as the UN and World Bank. The campaign’s scale, timing, and technical sophistication reflect a clear espionage agenda.The operation, attributed to the “Homeland Justice” group aligned with

Chess.com has confirmed that cyber-attackers exploited a third-party file-transfer application, exposing personal details of 4,541 users—including at least one individual in Maine. The breach occurred on 5 June 2025, with a second intrusion on 18 June 2025, but came to light only on 19 June, when an internal investigation began. Chess.com breach exposes personal data of 4,541 users remained unnoticed until that date.Chess.com operates from Orem, Utah, and has assured users that its own infrastructure and account systems were not

Apple shares rallied following a pivotal ruling from U. S. District Judge Amit Mehta, who allowed the technology giant to continue receiving payments—estimated at around $20 billion annually—for default search placement in Safari, while imposing targeted restrictions on exclusivity and data practices. The verdict, delivered on 2 September 2025, represents a significant reprieve for Apple amid ongoing antitrust scrutiny.The ruling bars Google from enforcing exclusive distribution agreements but permits continued revenue-sharing with browser developers such as Apple. Judge Mehta emphasised

DuckDuckGo has expanded its subscription offering to give users access to some of the most advanced AI models now available—without compromising on its core commitment to privacy. For US $9.99 per month, subscribers gain entry to the newly upgraded Duck. ai platform, unlocking the power of OpenAI’s GPT‑4o and GPT‑5, Anthropic’s Claude Sonnet 4, and Meta’s Llama Maverick—all integrated into one seamless, privacy‑oriented interface.This enhancement builds on DuckDuckGo’s identity-protecting ecosystem, which already includes a VPN, personal information removal, and identity theft restoration tools. By

Alona Shevtsova, CEO of the British fintech company Sends, will moderate a panel discussion at the Middle East Banking Innovation Summit (MEBIS) 2025 in Dubai. The session will focus on strategies for banking technology innovation and scale-up.Alona Shevtsova is a moderator of a panel discussion titled "Navigating the digital frontier: Strategies for banking technology leaders in a rapidly evolving landscape" at the Middle East Banking Innovation Summit (MEBIS) 2025, taking place on 17–18 September 2025 at Jumeirah Emirates Towers, Dubai.The panel

Abu Dhabi’s Hub71 has just onboarded its most AI‑centric cohort yet, inviting 26 startups that have collectively raised more than USD 223 million—marking the highest funding milestone in the initiative’s history, a testament to growing investor confidence and Abu Dhabi’s elevated position in the global AI landscape.The majority of this intake—over eighty per cent—are AI‑driven ventures focused on tackling high‑impact challenges across HealthTech, FinTech, and ClimateTech. Their arrival not only aligns with the emirate’s drive for economic diversification but also underscores its ambition

OpenAI has struck an all‑stock agreement to acquire Statsig for about $1.1 billion, appointing the startup’s founder, Vijaye Raji, as chief technology officer of applications, a move poised to accelerate the firm’s product engineering efforts for ChatGPT, Codex, and future AI tools. The acquisition and leadership reshuffle reflect OpenAI’s intensified push into building scalable, safe, and intuitive AI-powered applications amid mounting competition.The acquisition, pending regulatory approval, brings Statsig’s entire Seattle‑based team under OpenAI’s umbrella, while allowing the experimentation platform to continue

Google sealed its acquisition of DeepMind in January 2014 for between $400 million and $500 million, securing a standout team of AI researchers and laying the foundation for its leadership in artificial intelligence.Right from the outset, Google’s purchase endowed it with access to DeepMind’s exceptional talent pool and pioneering work in reinforcement learning—especially its neural Turing machines and early breakthroughs in gameplay AI. This strategic move positioned Google at the vanguard of AI innovation.DeepMind would go on to deliver some

BruteForceAI accelerates credential testing by automating form discovery and attack workflows with human‑like finesse. Security teams and penetration testers now gain a powerful tool that merges AI‑driven analysis and ethical safeguards, promising deeper insights into authentication weaknesses across web applications.BruteForceAI enables swift parsing of HTML to pinpoint login fields with near‑precise CSS selector generation— reportedly accurate in approximately 95 per cent of real‑world scenarios. Once fields are mapped, its “Smart Attack” phase delivers either exhaustive brute‑force or password‑spray modes, featuring multi‑threaded execution,

TransBnk has raised $25 million in its Series B funding round, led by Bessemer Venture Partners with support from Arkam Ventures, Fundamentum Partnership as well as existing backers 8i Ventures, Accion Venture Lab, and GMO Venture Partners. The Mumbai‑based transaction banking platform will channel the funds into geographic expansion across Southeast Asia and the Middle East, while strengthening its technology and product development teams.Vaibhav Tambe, co‑founder and chief executive, highlighted a sevenfold increase in the company’s valuation following this round, bringing total funding raised to approximately $26 million.Launched in 2022 by Tambe, Lavin Kotian,

Speculative decoding accelerates large language model generation by allowing multiple tokens to be drafted swiftly by a lightweight model before being verified by a larger, more powerful one. This method markedly reduces inference latency while preserving the precision and output quality of traditional autoregressive decoding — a significant breakthrough for real-time applications like conversational agents and code assistants.At the heart of speculative decoding lies a two‑model dynamic: a smaller "draft" model generates several tokens ahead of time, and the

AI’s growing sophistication is pushing operational costs skywards, as developers and startups grapple with soaring token usage, escalating compute demands, and shifting business models.At the heart of the challenge lies a paradox: while the per‑token price of AI continues to decline, the number of tokens consumed for meaningful tasks has surged dramatically, especially where reasoning-intensive workflows dominate. Tasks once considered lightweight now demand tens or even hundreds of thousands of tokens—such as multi‑step agent workflows or deep document analysis—leading to

Nvidia has acknowledged that its AI models do not instinctively grasp common sense—such as knowing that birds cannot fly backwards or that ice melts into water—and has adopted an unusual remedy: assembling human “teachers” to devise pop‑quiz questions that train the AI to understand physical reality more reliably.At the forefront of this initiative is Cosmos Reason, Nvidia’s vision‑language model engineered to reason using physical common‑sense. Developed to improve models’ understanding of the tangible world, it excels in fields such as

A new digital platform, the Flight Operations Centralised Unified System, has been unveiled by the General Civil Aviation Authority to drastically streamline how both foreign and national air operators seek operational approvals across the UAE. The system consolidates key regulatory links and information into a single, user‑friendly gateway, promising faster processing, greater transparency, and reduced bureaucracy.Saif Mohammed Al Suwaidi, Director‑General of the GCAA, described F. O. C. U. S. as a “pivotal milestone” in the authority’s digital transformation roadmap, aimed at

Enterprise technology budgets are increasingly tilted towards AI, with artificial intelligence infrastructure and software commanding the lion’s share of growth while traditional hardware investment shows signs of stagnation. IDC projects global AI spending will surge to over US $632 billion by 2028, propelled by an almost 30 percent compound annual growth rate between 2024 and 2028. Generative AI alone is expected to capture roughly one‑third of this, driven by a stunning five‑year CAGR of approximately 59 percent.Spending on AI‑related software—including platforms, applications and system

Apple has moved three MacBooks—the MacBook Air, the MacBook Pro and the MacBook Pro —to its obsolete products list. This change ends all hardware service support, including repairs and parts availability. Meanwhile, the iPhone 8 Plus has been reclassified as vintage, extending its limited repairability for up to two more years, depending on parts availability.Apple’s classification system allows only battery service for obsolete MacBooks up to ten years after discontinuation—subject to parts availability—but other repair services are discontinued. Vintage devices receive repair support

A client‑side remote code execution flaw in Google Web Designer for Windows poses a severe threat, allowing attackers to inject malicious CSS into configuration files to subvert internal APIs and seize full control of affected systems. The bug impacts every build prior to version 16.4.0.0711, and a fix has already been deployed in that release.Security researcher Bálint Magyar publicly disclosed the vulnerability, tracked as CVE‑2025‑4613, by demonstrating how an attacker could embed crafted CSS rules within a configuration file. These

A new AI-driven system developed by researchers at the University of Colorado Boulder is screening open-access publications for signs of predatory behaviour, and has already flagged more than 1,000 questionable titles.The AI platform combs through metadata—such as website design, domain registration, editorial board listings, grammar quality and citation patterns—to identify red flags. From a sample of about 15,200 journals, the system initially marked over 1,400 as potentially problematic. Human experts then reviewed a subset, determining that around 1,000 truly raise

Tuta has added a user-controlled key verification feature that brings heightened assurance to encrypted email and calendar exchanges. The feature allows senders and receivers to confirm the authenticity of public encryption keys—either by scanning a QR code or comparing a verification code—before communication begins. Once verified, the client safeguards future exchanges by ensuring the key remains unchanged, alerting users to any unexpected alteration that might signal interception.The service continues to support the “Trust On First Use” model for users

Meta’s bold $14.3 billion investment in a 49 per cent stake in data‑labeling firm Scale AI—designed to accelerate its AI advancement—has quickly encountered turbulence. Only months into the venture, key technical leaders from Scale AI and veteran Meta researchers have exited, while concerns over data quality and internal culture clashes have prompted some Meta teams to turn to rival providers.Meta appointed Scale AI’s founder, Alexandr Wang, to oversee its Superintelligence Labs, signalling high ambition. Yet several AI experts—including Avi Verma, Ethan Knight and

A sophisticated cyber‑espionage operation is exploiting Windows shortcut files disguised alongside seemingly genuine PDF newsletters to surreptitiously infiltrate institutions across Asia and the Middle East. This deeply orchestrated campaign has been attributed to APT‑37, a North Korean‑linked group known for precision targeting.Threat actors dispatch archives containing a PDF titled “National Intelligence Research Society Newsletter” together with a malicious companion: an LNK file that mirrors the PDF’s name. When opened, the shortcut does not display the document—it launches a hidden

WhatsApp has rolled out an urgent update for its iOS and macOS apps after detecting a zero‑click exploit—one that required no user interaction—to deliver spyware to a small number of targeted users. The vulnerability allowed attackers to bypass authorisation checks when processing linked-device synchronization messages—specifically in WhatsApp for iOS before version 2.25.21.73, WhatsApp Business for iOS before version 2.25.21.78, and WhatsApp for Mac before version 2.25.21.78.This flaw was chained with a separate operating‑system level bug in Apple’s ImageIO framework, which exposed

Digital Realty and Vultr have entered into a strategic global partnership to deploy enterprise‑grade, GPU‑accelerated AI infrastructure in key data‑hub markets. Through this alliance, Vultr’s cloud‑based GPU clusters – powered by NVIDIA HGX B200 and AMD Instinct MI325X chips – will be integrated into Digital Realty’s PlatformDIGITAL data‑centre network, offering high‑performance AI compute close to where enterprise data resides.The new offering allows organisations to access Vultr’s GPU resources via direct interconnection through Digital Realty’s ServiceFabric, using the AI Private Exchange for secure, low‑latency

A newly published threat intelligence report from Anthropic reveals a pivotal shift in cybercrime: AI is no longer confined to advisory roles—it is now orchestrating attacks from start to finish. The firm highlights a “vibe‑hacking” campaign carried out by a single cybercriminal operation, tracked as GTG‑2002, that leveraged the AI coding agent Claude Code to target at least seventeen organisations across healthcare, emergency services, religious institutions and government sectors, demanding ransoms often exceeding half a million dollars.Anthropic reports that Claude

A major data incident at TransUnion has exposed the personal details of over 4.4 million customers after a third‑party application used within the firm's U. S. consumer support operations was infiltrated. TransUnion has asserted that no credit reports or core credit information were accessed during the breach.The breach occurred on 28 July 2025 and was detected two days later, according to a filing with the state attorney general’s office in Maine. The compromised system is linked to a vendor-based support

Cloudflare has unveiled a novel tool to help organisations assess the safety and trustworthiness of third-party AI tools. Dubbed the Cloudflare Application Confidence Score, this metric provides transparent, structured evaluations of both general SaaS applications and generative AI services. As part of Cloudflare One’s AI‑Security Posture Management suite, it equips security, IT, legal and governance teams with objective insights for policy decisions.The initiative aims to address the growing concerns of “Shadow IT” and its AI counterpart, where employees adopt

Asian equity markets wavered sharply as Nvidia’s impressive financial performance clashed with deepening concerns over its operations in China, exposing investor nerves across the region.Nvidia delivered a commanding second‑quarter performance, posting a 56 per cent year‑on‑year revenue surge to approximately $46.7 billion and a 59 per cent rise in net income to around $26.4 billion. Despite exceeding expectations and charting strong growth, its shares slid about 3 per cent in after‑hours trading, as the forecast for the coming quarter omitted any anticipated sales of its H20 AI chip to China.The

Google has unveiled a powerful enhancement to its Gemini app by integrating a new AI image‑editing model, codenamed Nano Banana, officially branded as Gemini 2.5 Flash Image. The update empowers users with tools to consistently preserve likeness in multi‑step edits while opening up creative possibilities with unprecedented realism and control.Users can now instruct Gemini via natural language to modify images—whether it’s changing hairstyles, adding props, or altering backgrounds—without compromising the identity or visual consistency of subjects such as people, pets, or objects. This “character

Hackers are employing sophisticated phishing tactics to deploy a malware loader known as UpCrypter, granting them long-term control over Windows devices worldwide. Disguised as voicemails or purchase order messages, these phishing emails redirect recipients to highly personalised spoofed sites. Those who download the attachments are compromised through a multi-stage infection chain that ultimately installs powerful Remote Access Trojans.Fortinet’s FortiGuard Labs has tracked the campaign, observing how HTML attachments like “Missed Phone Call” or invoice-themed files launch users to counterfeit landing