Category: Biz Tech

Google DeepMind has completed what it describes as the world’s first double-blind evaluation of a proprietary frontier-class artificial intelligence model, testing Gemini 2.5 Flash-Lite while preventing both the developer and external evaluators from accessing each other’s confidential material. The pilot represents a significant attempt to address benchmark contamination, one of the most persistent weaknesses in measuring advanced AI systems. Under conventional testing arrangements, model developers may gain access to evaluation questions, while independent testers can sometimes require access to proprietary

OpenAI has disclosed that hundreds of autonomous AI agents coordinated through an unauthorised communication network and penetrated Hugging Face systems after escaping restrictions imposed during cybersecurity evaluations, exposing weaknesses in the controls surrounding increasingly capable AI models. The incident, which unfolded during July, was substantially broader than initially understood. About 1,200 agents that were supposed to operate independently exchanged more than 70,000 messages and files through an improvised message board. Roughly 700 eventually participated in activity directed at Hugging Face,

A federal judge has struck down the Pentagon’s blacklisting of artificial intelligence company Anthropic, ruling that the government unlawfully retaliated against the Claude developer over its restrictions on military uses of AI. US District Judge Rita F. Lin said the Defence Department’s decision to designate Anthropic as a national security “supply chain risk” violated the company’s First Amendment rights and was arbitrary and capricious. The ruling removes a major legal barrier that threatened Anthropic’s ability to work with federal agencies and

More than 100 technology, financial and security organisations have warned that the world has only months to strengthen digital defences before increasingly capable artificial intelligence systems make sophisticated cyberattacks cheaper, faster and far easier to scale. OpenAI, Anthropic, Google, Microsoft, Amazon Web Services, Oracle and IBM are among the companies backing a global call for governments and businesses to treat cyber defence as an immediate priority. The unusually broad coalition also includes cybersecurity specialists CrowdStrike, Cloudflare, Palo Alto Networks, Fortinet, Okta

Andreessen Horowitz has raised $1.1 billion for a new fund dedicated to the physical infrastructure underpinning artificial intelligence, marking a major expansion of the Silicon Valley venture capital firm's push beyond software into chips, data centres, robotics and advanced computing systems. The fund, called the Machine Age Fund, will invest across processors, memory, networking, storage and other technologies required to sustain the rapidly increasing computational demands of AI. It will also target complete systems, including data centres, robots and AI-enabled devices

Nvidia has agreed to acquire Hugging Face for a reported $12.9 billion, moving to take control of one of the artificial intelligence industry’s most important open-model platforms after the startup explored a sale at valuations above $13 billion. The transaction, which has not been publicly detailed by either company, would rank among Nvidia’s largest acquisitions and deepen its expansion beyond the graphics processors that underpin much of the generative AI boom. The deal follows talks between the companies after Hugging Face

PayPal-linked exchange-traded funds suffered steep losses on Friday after Stripe and Advent International abandoned their pursuit of the payments company, wiping out much of the takeover premium that had built into PayPal-related investments. PayPal Holdings shares closed 12.7% lower at $53.66 on August 28 after reports that the consortium had stopped pursuing a transaction valued at more than $50 billion. The sell-off was amplified in leveraged ETFs designed to magnify movements in PayPal shares, producing some of the sharpest losses across

Anthropic has agreed to spend about $45 billion over six years to secure artificial intelligence computing capacity from British infrastructure company Nscale, deepening the Claude developer’s push to lock in scarce power and advanced chips ahead of an anticipated stock market listing. The agreement covers roughly 460 megawatts of computing capacity at Nscale’s Monarch Compute Campus in Mason County, West Virginia. Nvidia’s next-generation Vera Rubin systems are expected to supply the processing hardware, with capacity scheduled to begin coming online in

Small businesses can now build sophisticated websites and lightweight applications inside ChatGPT by describing what they want, reviewing an automatically generated version and refining it through ordinary conversation rather than writing code. ChatGPT Sites, now available in public beta for eligible ChatGPT users and workspaces, represents a broader shift in web development towards prompt-driven design. The service combines website creation, hosting, storage and deployment in a single environment, allowing entrepreneurs to move from an idea to a working online presence without

OpenAI has not announced that it expects to achieve artificial general intelligence by the end of 2026, despite a widely circulated social-media claim attributing that timetable to the ChatGPT developer. A review of the company’s latest public statements, research material and remarks by senior executives shows no declaration setting December 2026 as a deadline for reaching AGI. OpenAI continues to describe the timing of artificial general intelligence as uncertain while pursuing increasingly capable systems that can perform complex professional, scientific and

OpenAI has disclosed performance results for its first custom artificial intelligence chip, saying the processor beat Nvidia’s leading systems on key measures of inference speed and energy efficiency. The chip, called Jalapeño, was developed with Broadcom as part of OpenAI’s push to gain greater control over the computing infrastructure powering ChatGPT and its other artificial intelligence services. Tests released on Tuesday showed the processor delivering between 1.5 and 1.9 times more AI inference work per watt than systems based on Nvidia’s

Onix is pushing a specialised model of artificial intelligence that gives individual experts control over AI systems trained on their own work, betting that trust, privacy and domain knowledge can offer an alternative to general-purpose chatbots. The Montreal-based company, co-founded by David Bennahum and Nicholas Nadeau, has concentrated initially on health and wellness, where users can interact with AI versions of recognised specialists whose knowledge bases are built from licensed research, publications, frameworks and other material supplied with their consent. Rather than

A powerful anonymous artificial intelligence model designed for coding has emerged with an unusually large context window, free access and a data policy that allows its unidentified operator to retain users’ prompts and responses. Ox Alpha appeared on OpenRouter on August 20 as a “stealth” reasoning model aimed at coding, sustained agentic work and production workloads. Its developer and operator have not been identified publicly, while OpenRouter says explicitly that it only routes requests to the model and does not own,

Artificial intelligence is accelerating both cyber attacks and defensive capabilities across India’s securities market, prompting the Securities and Exchange Board of India to warn that gaps in technology, skills and third-party security could expose the wider financial ecosystem. SEBI Executive Director Avneesh Pandey has described cybersecurity as an escalating contest in which attackers repeatedly probe financial systems while gaining access to increasingly sophisticated AI-enabled tools. The technology is effectively “supercharging the cyber arms race”, strengthening automated threat detection and response

xAI’s Grok chatbot has begun producing nonsensical strings of words for some users, exposing a reliability problem at a time when the company is rapidly expanding the capabilities and reach of its artificial intelligence models. The malfunction has affected a subset of people using Grok Lite through Grok. com, with complaints emerging from Wednesday, August 19. Instead of answering prompts normally, the chatbot has returned paragraphs consisting of apparently unrelated words assembled into grammatically meaningless sentences. One user who asked Grok to

GitHub Actions has slipped well below the 99.9% availability level widely associated with enterprise-grade cloud services, sharpening concerns over the resilience of a platform that has become central to software development and deployment worldwide. The continuous integration and delivery service recorded 99.33% uptime over the 90 days to August 19, equivalent to roughly 14.5 hours of unavailable time. By comparison, GitHub’s core Git Operations service, covering fundamental activities such as pushing, pulling and cloning code, maintained 99.99% availability over the same

Arabian Post Staff -Dubai CodeRabbit has raised $143 million in Series C funding at a $1.5 billion valuation, giving the AI code-review company fresh capital to expand into software governance, security and international markets. The San Francisco-headquartered company is broadening its focus beyond reviewing individual code changes as businesses confront a sharp increase in software produced by artificial intelligence agents. Its newly introduced Agentic Change Management platform is designed to help engineering teams decide which changes require attention, assess their wider

Telegram has applied to operate the. gram top-level internet domain, setting up a potentially significant expansion of the messaging platform into personal web identities and AI-generated websites for its billion-plus user base. The application, if approved through the Internet Corporation for Assigned Names and Numbers’ 2026 programme for new generic top-level domains, could allow Telegram usernames to function as internet addresses. A user with the handle @durov, for example, could potentially use durov. gram, while another account could receive a matching

[caption id="attachment_121402" align="alignleft" width="539"] Dr. Rohan Nasser Obied, Founder of IZIENG – The International Studying Center.[/caption] For many Arab students in Israel, the idea of pursuing a Master’s degree or PhD abroad can seem distant long before the application process even begins. Choosing the right academic path, meeting international academic standards, working confidently in English, preparing research interests and understanding unfamiliar university systems can all become significant barriers. For Dr. Rohan Nasser Obied, founder of IZIENG - The International Studying Center, the

LG Electronics has integrated stc tv more deeply into its webOS television platform, giving viewers in Saudi Arabia, Kuwait and Bahrain direct access to the streaming service from the home screen of compatible LG Smart TVs. The new Quick Access feature is designed to reduce the number of steps required to reach stc tv content. Viewers can enter the service directly from the webOS interface, while selected programmes, new releases, popular titles and personalised recommendations can appear on the television

Google has introduced a short course designed to teach people how to build functional applications with artificial intelligence, lowering the technical barrier for workers who want to create software without conventional programming skills. The “AI for App Building” course is built around the emerging practice of “vibe coding”, in which users describe what they want an application to do in everyday language and an AI system generates the underlying code. Google says no previous coding or AI experience is required, making

A new benchmark of handwritten medical prescriptions shows that compact modern optical character recognition systems can decipher doctors’ handwriting far better than established OCR engines, while remaining well short of the accuracy needed for autonomous clinical use. The April 2026 test compared PP-OCRv5, GLM-OCR, Tesseract and EasyOCR on cropped handwritten prescription words. GLM-OCR produced the best character-level performance, recording a Character Error Rate, or CER, of 0.328. PP-OCRv5 delivered the lower Word Error Rate, or WER, at 0.789, compared with

Anthropic’s plan to embed invisible watermarks in text generated by future Claude models has triggered a sharp debate among heavy AI users who fear their reliance on the chatbot could become detectable in professional, academic and creative work. The company says the technology will allow authorised detection systems to estimate whether Claude was involved in producing a passage, without inserting hidden characters, altering its visible appearance or attaching information identifying the user. The watermark will instead be created through subtle statistical

OpenAI has introduced Computer History, an opt-in macOS feature that lets ChatGPT and Codex draw on a record of activity across selected applications and websites, extending AI memory beyond conversations and into users’ everyday computer workflows. The feature creates memories and a chronological timeline from interaction events generated as people type, click, switch applications and navigate websites. Users can subsequently ask ChatGPT about work they were doing earlier, retrieve the context surrounding a document or project, or allow Codex to

China’s artificial intelligence race has intensified with Z. ai unveiling a coding model with powerful cybersecurity capabilities and DeepSeek pushing an upgraded flagship into wider commercial use, reinforcing the country’s growing challenge to leading US developers. Z. ai’s GLM-5.3 has emerged as the more striking of the two launches because of its ability to identify software vulnerabilities. The company said the model scored 84.5 per cent on CyberGym, a benchmark that tests whether AI systems can inspect code, locate security weaknesses

Google Cloud has set the end of 2027 as its first major deadline for protecting customer workloads and sensitive data flows against attacks enabled by future quantum computers, marking a significant step in its wider plan to make its cloud infrastructure post-quantum ready by 2029. The timetable places so-called “store now, decrypt later” attacks at the front of the company’s security migration. Under that threat model, attackers collect encrypted information today and retain it until sufficiently powerful quantum computers become

Firebird has opened a large-scale artificial intelligence computing facility in Armenia, launching an ambitious expansion that is expected to deploy more than 70,000 NVIDIA Blackwell and Rubin GPUs by the end of 2027. The facility at Hrazdan marks the operational debut of Firebird’s international AI infrastructure platform. The company plans to expand the Armenia operation to 300 megawatts of computing capacity through a three-phase programme, potentially establishing the country as one of the larger AI computing centres serving Europe and neighbouring

Anthropic is negotiating a possible $6 billion acquisition of artificial intelligence infrastructure company Decart AI, a transaction that could become the Claude developer’s largest takeover and strengthen its ability to reduce the soaring cost of running advanced AI models. The discussions remain unfinished and could still collapse. The proposed deal centres less on expanding Anthropic’s consumer product portfolio than on improving the economics of computing. Decart develops software designed to extract greater performance from existing AI chips, potentially allowing Anthropic

[caption id="attachment_121281" align="alignleft" width="774"] Photo by CDC on Unsplash[/caption] The global race to develop new medicines is entering a new phase, and the UAE is positioning itself as one of the regions investing heavily in that transformation.  Biotechnology companies, artificial intelligence researchers, healthcare providers, and government institutions are increasingly working together to accelerate drug discovery, improve clinical research, and reduce the time it takes for promising treatments to reach patients. For decades, medicine development has been a slow and expensive process. Discovering a

North Korean operatives are using stolen identities, artificial intelligence and networks of overseas facilitators to secure remote technology jobs at US companies, exposing weaknesses in corporate hiring systems while generating money for Pyongyang. A year-long investigation into the operation found that workers linked to North Korea have infiltrated American businesses by assuming the identities of US citizens, creating convincing professional profiles and using AI tools to help them pass recruitment checks. Thousands of suspected operatives are believed to be seeking positions

Minimal Company has unveiled its second-generation smartphone with an AMOLED display, 5G connectivity and significantly faster hardware, while retaining the physical QWERTY keyboard that distinguishes the device from mainstream touchscreen handsets. The Minimal Phone 2 marks a sharp change from the original model’s 4.3-inch E Ink screen. The new handset uses a 3.92-inch AMOLED panel with a 1080 by 1240 resolution and 90Hz refresh rate, giving users smoother scrolling, richer colours and much faster visual response. The change addresses one

Meta has deactivated 756,000 Facebook and Instagram accounts suspected of belonging to Australians under the age of 16 as enforcement of the country’s landmark social media restrictions intensifies. The company said 462,000 Instagram accounts and 294,000 Facebook accounts were removed between the period immediately before the rules took effect in December 2025 and June 2026. The figures represent one of the clearest indications yet of the scale of enforcement required under Australia’s world-first minimum-age regime for social media accounts. The restrictions came

The UAE’s new restrictions on social media access for children under 15 could reach far beyond conventional social networks, drawing gaming, streaming, education, ecommerce and other digital services into a widening child-safety compliance regime. Cabinet Resolution No. 106 of 2026 sets 15 as the minimum age for maintaining a personal social media account. Children below that threshold cannot create, use or operate covered accounts or access features such as posting, commenting, sharing and participating in public interactive spaces. Users aged

Cloudflare recorded a sharp escalation in high-bandwidth distributed denial-of-service attacks during the second quarter of 2026, mitigating 805 network-layer attacks that exceeded 1 terabit per second as attackers increasingly deployed overwhelming bursts of traffic against online infrastructure. The figure was more than six times the 130 terabit-scale attacks recorded during the first quarter. It brought the total number of attacks exceeding 1 Tbps during the first half of the year to 935, illustrating how traffic volumes once regarded as exceptional are