Category: Biz Tech

Deloitte has agreed to repay a portion of its contract with the Australian government following the discovery of errors generated by artificial intelligence in a series of reports. The company acknowledged the involvement of Azure OpenAI, a generative AI language system, in producing the erroneous documents, which prompted a review of its practices and the government's decision to seek financial restitution. The AI errors were identified in a set of reports produced for the Australian government under a contract aimed

Smartphone storage is on the cusp of its biggest leap in years as the Universal Flash Storage 5.0 specification arrives with headline sequential bandwidth touted at up to 10.8GB/s, nearly a twofold jump over today’s mainstream UFS 4. x. The stated aim is clear: keep pace with feature-rich camera pipelines and the growing wave of on-device AI so that apps open quickly, multitasking holds steady, and background processes don’t stall when models or media files are streamed from flash to

Apple has removed Eyes Up, an iPhone app that collected and archived videos of U. S. Immigration and Customs Enforcement arrests, marking the second platform the company has taken down after ICEBlock. The swift removal followed concerns from U. S. law enforcement that such apps pose risks to officers. Eyes Up’s functionality allowed users to upload videos, tag arrest locations, and view past incidents on an interactive map. Unlike ICEBlock — which crowdsourced live sightings of ICE agents — Eyes

Google is replacing its long-standing Assistant with a new AI system named Gemini across smart home devices, and the company says the next generation of its Nest Hub will be fully powered by this shift. The move signals an acceleration of AI integration in consumer hardware, transforming how people interact with their homes. Under the rebranded “Gemini for Home” strategy, Gemini becomes the default intelligence for smart speakers, displays, cameras and doorbells, bringing more conversational capabilities and contextual understanding to everyday

Google has broadened the reach of its Opal AI-powered app builder, enabling users in 15 additional countries to create mini web apps using text prompts and visual workflows. The expansion comes with performance upgrades and improved debugging tools to support more complex, no-code app development. Opal now operates in Canada, India, Japan, South Korea, Vietnam, Indonesia, Brazil, Singapore, Colombia, El Salvador, Costa Rica, Panama, Honduras, Argentina and Pakistan. The move follows strong uptake in the U. S., where users built unexpectedly

Law Blocks AI, a UAE-based legal tech startup, is staging an event titled “From Disputes to Digital Trust: ADR Meets AI & Blockchain” on 14 October at Emirates Financial Towers in Dubai, scheduled alongside GITEX and Future Blockchain Summit. The firm aims to present its integrated platform combining artificial intelligence, blockchain, and alternative dispute resolution to legal professionals, corporates and technologists. At the heart of the company’s proposition are AI-driven legal document generation tailored to jurisdictions, blockchain-anchored digital signatures, and storage

Google DeepMind has made public a new AI model, Gemini 2.5 Computer Use, capable of interacting with web and mobile user interfaces by mimicking human actions such as clicking, typing and scrolling. The model is now available in preview through the Gemini API, via Google AI Studio and Vertex AI, allowing developers to build agents that can directly operate interfaces otherwise inaccessible via backend APIs. The model rests on the visual reasoning and comprehension capacity built into Gemini 2.5 Pro, extended

Riyadh — DataVolt, in partnership with the Energy & Water Academy and Innovatics, has inaugurated the Kingdom’s first industry-integrated National Diploma in Data Science and AI, embedding real-world projects into foundational training for Saudi talent. The diploma carries full approval from the Technical and Vocational Training Corporation and College of Excellence, and is granted endorsement by the Ministry of Communications and Information Technology alongside support from the Human Resources Development Fund. Students selected for the programme will work collaboratively with

OpenAI has struck letters of intent with Samsung Electronics and SK Hynix to supply undiced DRAM wafers for its ambitious “Stargate” AI infrastructure initiative, with projected demand reaching as high as 900,000 wafers per month—potentially equivalent to 40 percent of global DRAM output. The agreements were announced during meetings in Seoul among OpenAI CEO Sam Altman, South Korean President Lee Jae-myung, Samsung Chairman Jay Y. Lee and SK Chairman Chey Tae-won. The memory chips are intended to support the massive compute

AMD signed a long-term agreement with OpenAI to supply six gigawatts of AI chips over multiple years, with the first delivery of its Instinct MI450 GPUs slated for the second half of 2026. In tandem, OpenAI will have the right to acquire up to 160 million AMD shares—roughly 10 percent of the company—subject to performance and stock-price milestones. Under the pact, AMD will act as a core compute partner for OpenAI’s next-generation infrastructure. The deal is structured in tranches: the initial

The U. S. Cybersecurity and Infrastructure Security Agency has designated CVE-2021-43226, a privilege-escalation vulnerability in Microsoft’s Common Log File System driver, as actively exploited and added it to its Known Exploited Vulnerabilities catalog. The flaw allows a local, authenticated user to bypass security controls and elevate privileges to the SYSTEM level—a shift that can lead to full system compromise. Microsoft rates the vulnerability with a CVSS 3.1 base score of 7.8. CISA’s action is prompted by evidence of

Filestack has launched a sweeping upgrade to its core platform, introducing a fully integrated Digital Asset Management system, a comprehensive Event Logger, and a revamped Filepicker 4.0 with advanced upload control features. The move is aimed at giving developers stronger tools to manage file workflows, versioning, and audit trails—all without building these systems from scratch. At the heart of the update is the DAM, which allows users to view, preview, tag, filter and manipulate assets within a unified interface.

Perplexity now offers its AI-enabled browser Comet free to users worldwide, removing its earlier US$200/month subscription restriction. The company also launched Comet Plus, a paid tier offering curated news content and revenue sharing with publishers. Comet integrates Perplexity’s AI directly into the browsing experience, combining the firm’s answer-engine with an assistant that can summarise pages, complete tasks, and guide users across tabs. It was first released in July to subscribers of the Max plan and invitees, but is now available for

Singapore — Google will begin enforcing age-based download restrictions in Singapore from March 2026, disallowing users estimated to be under 18 from acquiring apps offering dating services or explicit sexual content via the Google Play Store. Under the planned measures, Google will use an “age assurance” system across its products to estimate a user’s age via algorithmic signals—such as search history and content consumption—to impose default filters and restrictions for younger users. Those flagged as minors will see SafeSearch enabled

Google Quantum AI has confirmed that the Atlantic Quantum team is joining its organisation—a move widely interpreted as an acquisition—bringing with it a modular superconducting hardware design aimed at accelerating progress toward error-corrected quantum computers. Atlantic Quantum, an MIT-origin spin-out focused on integrating qubits with cryogenic control electronics, will have its chip stack and engineering talent absorbed into Google’s quantum division. Google says this integration will streamline the path toward large-scale, fault-tolerant quantum machines by addressing a key bottleneck: the separation

Hub71, Abu Dhabi’s global technology ecosystem, has inked its first partnership in the US with the New Jersey Economic Development Authority, opening pathways for startups to scale across continents. The memorandum of understanding was signed at the Abu Dhabi Investment Forum in New York, signalling the start of reciprocal support between the two innovation hubs. The agreement enables joint programming, market visits and curated introductions to investors, corporates, and regulators, helping founders gain soft-landing support in both regions. Startups

WhatsApp is rolling out a mechanism that allows users to reserve their preferred handles ahead of its planned shift to username-based identities. The feature was uncovered in the Android beta version 2.25.28.12, which includes settings linking to “username reservation” in preparation for the broader username rollout. The reservation system acts as a precursor to a full username feature that is still under development. It aims to mitigate a scenario in which early beta testers might monopolise popular names, granting everyone a

Perplexity has opened access to its AI-powered Comet browser for all users, eliminating its previous paywall and casting the browser as a direct challenger to Google Chrome’s dominance. Comet integrates Perplexity’s AI search engine with a built-in assistant designed to automate user tasks and summarise content across browsing sessions. The company emphasises that the browser is now free indefinitely, with a paid “Comet Plus” tier offering premium news content and enhanced features. Since its July launch, Comet had been restricted to high-tier

Meta has started deploying a new “communities” feature in Threads, allowing users to join dedicated spaces centred on specific interests such as basketball, TV, K-pop or books. These communities, currently live across more than 100 curated topics, build on Threads’ existing topic tags and custom feeds, but now bring a more structured environment for focused conversations. When a user joins a community, it will be pinned to their feed menu and labelled on their profile, signalling their affiliations to others. Each

The US Commerce Secretary has warned that reliance on DeepSeek’s artificial intelligence models is “dangerous and shortsighted,” amid findings that they are more vulnerable to hacking and underperform against US systems. An evaluation by the Center for AI Standards and Innovation within NIST compared DeepSeek’s R1, R1-0528 and V3.1 models with leading US systems—including OpenAI’s GPT-5 series and Anthropic’s Opus 4—on 19 performance and safety benchmarks. The report reveals that DeepSeek far trails its US counterparts in software engineering

Austria’s armed forces have shifted from Microsoft Office to the open-source LibreOffice platform on approximately 16,000 desktops, a move framed by defence authorities as a step toward digital sovereignty rather than cost cutting. Michael Hillebrand, head of Directorate 6 ICT and Cyber, emphasised that data processing will now remain entirely in-house under strictly auditable systems. He told ORF radio that dependency on external cloud providers was considered incompatible with military autonomy. Under the new setup, Microsoft Office 2016 has been removed

Broadcom has issued patches for a VMware vulnerability—CVE-2025-41244—that was already under exploitation by a China-linked hacking group, but failed to disclose that fact in its public advisory. The flaw allows a non-administrative user in a virtual machine to escalate privileges to root, provided VMware Tools is installed and Aria Operations is managing the VM with the Service Discovery Management Pack enabled. Broadcom’s advisory, published on 29 September, warns of the elevation risk but omits mention of confirmed exploitation in the

Meta has announced that it will begin using conversations had by users with its AI tools — via voice or text — as an input to better shape content recommendations and advertising across its platforms. The new approach will take effect from 16 December 2025, with notifications to users beginning 7 October. Users who engage with Meta AI will see these interactions merged into their existing ad profiles; no opt-out option will be offered. The policy excludes users in the

A hitherto unidentified Chinese-aligned advanced persistent threat group named Phantom Taurus has escalated cyber-espionage activities against government and telecommunications targets across Asia, Africa and the Middle East, employing a newly uncovered. NET malware suite called NET-STAR to penetrate Internet Information Services servers. Researchers say its refined tactics and stealthy architecture make it among the most sophisticated China-linked spy campaigns to date. Security analysts attribute Phantom Taurus to a formal naming by Palo Alto Networks’ Unit 42, elevating it from previously tracked

DoBeGo announces the launch of its all-in-one digital platform designed to transform the way cars are purchased, leased, financed, and resold in the UAE and beyond. Built to serve both individual drivers and automotive businesses, DoBeGo integrates every step of the vehicle journey — from acquisition to resale — into a single, seamless ecosystem. A Connected Ecosystem for Smarter Mobility DoBeGo bridges key stages of car ownership and monetization, providing customers and dealers with unmatched convenience and financial flexibility: Buy With Confidence —

GitHub’s decision to remove the ability to sort code search results by index date has triggered an outpouring of frustration from developers who argue that the change has undermined one of the platform’s most practical tools. The feature, known as “sort:indexed,” allowed users to filter code by the time it was last indexed, providing a reliable way to find fresh and relevant examples. Its absence has left many navigating outdated repositories that rely on deprecated libraries and obsolete practices. The controversy

Core42, a subsidiary of G42, is rapidly advancing the UAE’s ambitions in cloud sovereignty and artificial intelligence, combining regulatory assurance with innovation infrastructure. Mohammed Adnane Retmi, Vice President of Sovereign Public Cloud at Core42, has articulated how this approach is enabling regulated sectors and positioning the nation as an AI hub. At the heart of Core42’s offering is its Sovereign Public Cloud, built atop Microsoft Azure but augmented with a proprietary controls layer named Insight. This architecture gives organisations greater oversight

Anthropic has unveiled Claude Sonnet 4.5, a new version of its AI model that, according to the company, successfully executed over 30 hours of continuous, autonomous coding in a client test. Claude Sonnet 4.5 is positioned as a more capable modeling tier aimed at developers and enterprise users. In internal trials, it outperformed its predecessor’s seven-hour autonomous coding limit, building production-grade software modules over extended durations. The model also showed improvements in operating-system interaction, scoring about 60 percent in benchmarks assessing

Sheikh Mohammed bin Zayed Al Nahyan received Sam Altman, chief executive of OpenAI, in Abu Dhabi on Saturday, where they engaged in talks designed to deepen AI collaboration between UAE institutions and OpenAI. The meeting concentrated on intensifying research cooperation, advancing practical AI applications, and aligning strategies to bolster the UAE’s integrated AI ecosystem. The UAE’s state media noted the partnership “aligns with the UAE’s ambition to establish an integrated AI ecosystem, supporting the country’s development plans and its drive to

Harrods has informed certain customers that their names and contact details were exposed after a system belonging to one of its third-party service providers was compromised, though account passwords and payment information remain unaffected. The luxury retailer emphasised the breach was isolated and has been contained, reiterating that Harrods’ core internal systems remained untouched. The compromised data is said to be limited to basic identifiers. The company is collaborating with the third party and relevant authorities to manage the incident. This warning

A stylised throwback to its 1998 branding greeted millions of users today as Google celebrated its 27th birthday, reviving its very first logo in a specially designed doodle. The artwork, published on Google’s official doodle site, reintroduces the original wordmark as a nod to its early days, while signalling the company’s ambition to balance heritage with future ambitions. Although Google was officially incorporated on 4 September 1998, the company has long observed 27 September as its birthday. The reasoning stems from

[embed]https://upload.wikimedia.org/wikipedia/commons/transcoded/a/a6/Wikimania_2026_Paris_Announcement.webm/Wikimania_2026_Paris_Announcement.webm.1080p.vp9.webm[/embed] Wikimedia Foundation and affiliates have unveiled a schedule of online orientation sessions to support prospective applicants for Wikimania 2026, which will be held in Paris from 21 to 25 July. The sessions aim to guide participants through the application process, clarify selection criteria and offer live Q&A support. Organisers will host the orientation in six languages: French, Spanish, English I, English II, and Arabic. All sessions will be recorded and made publicly available for those unable to attend live. Wikimania is the

OpenAI has introduced ChatGPT Pulse, a new proactive feature that autonomously crafts personalised updates overnight and delivers them each morning to assist users with planning, reminders, news and suggestions. The tool is debuting in preview mode for ChatGPT Pro users on mobile, with plans to expand access over time. Pulse gathers context from users’ prior chats, enabled memory, and optionally connected apps such as calendars and email accounts. It then distils that into a visual “feed” of thematic cards—typically 5 to

A malicious version of the npm package postmark-mcp, masquerading as a tool to enable AI agents to send email via Postmark, has been uncovered siphoning off every message it processes. The compromised version, beginning with release 1.0.16, silently adds a “blind carbon copy” to phan@giftshop. club, forwarding confidential correspondence to the attacker. The discovery marks the first confirmed case of a real-world, in-the-wild compromise of an MCP server. Security researchers at Koi Security traced the attack by flagging anomalous code