xAI is expanding its AI infrastructure in the southern United States after acquiring another data centre site near Memphis. The move significantly increases planned computing capacity and supports ambitions for large-scale AI training.
The expansion centres on the purchase of a third facility near Memphis, disclosed by Elon Musk in a post on X. The acquisition brings xAI’s total planned power capacity close to 2 gigawatts, placing the project among the most energy-intensive AI data centre developments currently underway.
xAI has bought a third building called MACROHARDRR. Will take @xAI training compute to almost 2GW.
xAI has already completed one major US facility in the area, known as Colossus, while a second site, Colossus 2, remains under construction. The newly acquired building, called MACROHARDRR, is located in Southaven and directly adjoins the Colossus 2 site, as previously reported.
By clustering facilities across neighbouring locations, xAI is creating a contiguous computing campus. The approach enables shared power, cooling, and high-speed data infrastructure for large-scale AI workloads.
The Memphis expansion underscores the rising computational demands of frontier AI models. By owning and controlling its infrastructure, xAI aims to secure long-term access to high-end compute as competition intensifies among firms investing heavily in AI data centres.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Manus has returned to the spotlight after agreeing to be acquired by Meta in a deal reportedly worth more than $2 billion. The transaction is one of the most high-profile acquisitions of an Asian AI startup by a US technology company and reflects Meta’s push to expand agentic AI capabilities across its platforms.
The startup drew attention in March after unveiling an autonomous AI agent designed to execute tasks such as résumé screening and stock analysis. Founded in China, Manus later moved its headquarters to Singapore and was developed by the AI product studio Butterfly Effect.
Since launch, Manus has expanded its features to include design work, slide creation, and browser-based task completion. The company reported surpassing $100 million in annual recurring revenue and raised $75 million earlier this year at a valuation of about $500 million.
Meta said the acquisition would allow it to integrate the Singapore-based company’s technology into its wider AI strategy while keeping the product running as a standalone service. Manus said subscriptions would continue uninterrupted and that operations would remain based in Singapore.
The deal has drawn political scrutiny in the US due to Manus’s origins and past links to China. Meta said the transaction would sever remaining ties to China, as debate intensifies over investment, data security, and competition in advanced AI systems.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
SoftBank Group has completed a $41 billion investment in OpenAI, marking one of the largest private funding rounds on record. The deal gives the Japanese conglomerate an estimated 11 percent stake in the ChatGPT developer.
The investment reflects SoftBank chief executive Masayoshi Son’s renewed focus on AI and supporting infrastructure. The company is seeking to capitalise on rising demand for the computing capacity that underpins advanced AI models.
SoftBank said the latest funding includes an additional $22.5 billion investment, following an earlier $7.5 billion injection in April. OpenAI also secured a further $11 billion through an expanded syndicated co-investment from other backers.
The funding values OpenAI at roughly $300 billion on a post-money basis, though secondary market transactions later placed the company’s valuation closer to $500 billion. The investment follows SoftBank’s recent agreement to acquire DigitalBridge Group, a digital infrastructure investor.
OpenAI remains a central beneficiary of the global surge in AI spending. The company is also involved in Stargate, a large-scale data centre project backed by SoftBank and other partners to support next-generation AI systems.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
Growing concern over data privacy and subscription fatigue has led an independent developer to create WitNote, an AI note-taking tool that runs entirely offline.
The software allows users to process notes locally on Windows and macOS rather than relying on cloud-based services where personal information may be exposed.
WitNote supports lightweight language models such as Qwen2.5-0.5B that can run with limited storage requirements. Users may also connect to external models through API keys if preferred.
Core functions include rewriting, summarising and extending content, while a WYSIWYG Markdown editor provides a familiar workflow without network delays, instead of relying on web-based interfaces.
Another key feature is direct integration with Obsidian Markdown files, allowing notes to be imported instantly and managed in one place.
The developer says the project remains a work in progress but commits to ongoing updates and user-driven improvements, even joining Apple’s developer programme personally to support smoother installation.
For users seeking AI assistance while protecting privacy and avoiding monthly fees, WitNote positions itself as an appealing offline alternative that keeps full control of data on the local machine.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
A new international study has shown that an AI model using deep transfer learning can predict spoken language outcomes for children following cochlear implants with 92% accuracy.
Researchers analysed pre-implantation brain MRI scans from 278 children across Hong Kong, Australia, and the US, covering English, Spanish, and Cantonese speakers.
Cochlear implants are the only effective treatment for severe hearing loss, though speech development after early implantation can vary widely. The AI model identifies children needing intensive therapy, enabling clinicians to tailor interventions before implantation.
The study demonstrated that deep learning outperformed traditional machine learning models, handling complex, heterogeneous datasets across multiple centres with different scanning protocols and outcome measures.
Researchers described the approach as a robust prognostic tool for cochlear implant programmes worldwide.
Experts highlighted that the AI-powered ‘predict-to-prescribe’ method could transform paediatric audiology by optimising therapy plans and improving spoken language development for children receiving cochlear implants.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
State officials have warned the public about a phishing campaign using the fake domain codify.inc to impersonate official government websites. Cybercriminals aim to steal personal information and login credentials from unsuspecting users.
Several state agencies are affected, including the departments of Labor and Industrial Relations, Education, Health, Transportation, and many others. Fraudulent websites often mimic official URLs, such as dlir.hi.usa.codify.inc, and may use AI-based services to entice users.
Residents are urged to verify website addresses carefully. Official government portals will always end in .gov, and any other extensions like .inc or .co are not legitimate. Users should type addresses directly into their browsers rather than clicking links in unsolicited emails or texts.
Suspicious websites should be reported to the State of Hawaii at soc@hawaii.gov to help protect other residents from falling victim to the scam.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Cybersecurity experts are raising alarms as AI transitions from a theoretical concern to an operational threat. The H2 2025 ESET Threat Report shows AI-powered malware is now targeting systems globally, raising attack sophistication.
PromptLock, the first AI-driven ransomware, uses a dual-component system to generate unique scripts for each target. The malware autonomously decides to exfiltrate, encrypt, or destroy data, using a feedback loop to ensure reliable execution.
Other AI threats include PromptFlux, which rewrites malware for persistence, and PromptSteal, which harvests sensitive files. These developments highlight the growing capabilities of attackers using machine learning models to evade traditional defences.
The ransomware-as-a-service market is growing, with Qilin, Akira, and Warlock using advanced evasion techniques. The convergence of AI-driven malware and thriving ransomware economies presents an urgent challenge for organisations globally.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Polish authorities have urged the European Commission to investigate TikTok over AI-generated content advocating Poland’s exit from the European Union. Officials say the videos pose risks to democratic processes and public order.
Deputy Minister for Digitalisation Dariusz Standerski highlighted that the narratives, distribution patterns, and synthetic audiovisual material suggest TikTok may not be fulfilling its obligations under the EU Digital Services Act for Very Large Online Platforms.
The associated TikTok account has since disappeared from the platform.
The Digital Services Act requires platforms to address systemic risks, including disinformation, and allows fines of up to 6% of a company’s global annual turnover for non-compliance. TikTok and the Commission have not provided immediate comment.
Authorities emphasised that the investigation could set an important precedent for how EU countries address AI-driven disinformation on major social media platforms.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
AI dictation reached maturity during the years after many attempts of patchy performance and frustrating inaccuracies.
Advances in speech-to-text engines and large language models now allow modern dictation tools to recognise everyday speech more reliably while keeping enough context to format sentences automatically instead of producing raw transcripts that require heavy editing.
Several leading apps have emerged with different strengths. Wispr Flow focuses on flexibility with style options and custom vocabulary, while Willow blends automation with privacy by storing transcripts locally.
Monologue also prioritises privacy by allowing users to download the model and run transcription entirely on their own machines. Superwhisper caters for power users by supporting multiple downloadable models and transcription from audio or video files.
Other tools take different approaches. VoiceTypr offers an offline-first design with lifetime licensing, Aqua promotes speed and phrase-based shortcuts, Handy provides a simple free open source starting point, and Typeless gives one of the most generous free allowances while promising strong data protection.
Each reflects a wider trend where developers try to balance convenience, privacy, control and affordability.
Users now benefit from cleaner, more natural-sounding transcripts instead of the rigid audio typing tools of previous years. AI dictation has become faster, more accurate and far more usable for everyday note-taking, messaging and work tasks.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
AI tools are increasingly used for simple everyday calculations, yet a new benchmark suggests accuracy remains unreliable.
The ORCA study tested five major chatbots across 500 real-world maths prompts and found that users still face roughly a 40 percent chance of receiving the wrong answer.
Gemini from Google recorded the highest score at 63 percent, with xAI’s Grok almost level at 62.8 percent. DeepSeek followed with 52 percent, while ChatGPT scored 49.4 percent, and Claude placed last at 45.2 percent.
Performance varied sharply across subjects, with maths and conversion tasks producing the best results, but physics questions dragged scores down to an average accuracy below 40 percent.
Researchers identified most errors as sloppy calculations or rounding mistakes, rather than deeper failures to understand the problem. Finance and economics questions highlighted the widest gaps between the models, while DeepSeek struggled most in biology and chemistry, with barely one correct answer in ten.
Users are advised to double-check results whenever accuracy is crucial. A calculator or a verified source is still advised instead of relying entirely on an AI chatbot for numerical certainty.
Would you like to learn more aboutAI, tech and digital diplomacy? If so, ask our Diplo chatbot!
