Development Archives | Digital Watch Observatory

NVIDIA launches AI video detector with 92% accuracy

NVIDIA has introduced Synthetic Video Detector, an AI-powered microservice designed to help media organisations identify potentially AI-generated video.

The system analyses footage frame by frame and produces a probability score indicating whether it contains synthetic content.

Editorial teams can use the result to prioritise clips for review, quarantine suspicious footage or escalate it for more detailed verification.

NVIDIA said the tool is intended to supplement established fact-checking and forensic practices rather than determine authenticity on its own.

In company testing, the detector achieved up to 92% accuracy on uncompressed video. Accuracy declined to 87% at 15% compression and 82% at 50% compression.

The system can process 1080p video in as little as 22 milliseconds on NVIDIA RTX systems and around 30 milliseconds on NVIDIA L40 GPUs.

Its architecture uses an ensemble of DINOv2 and DINOv3 vision transformer backbones and focuses on artefacts introduced by generative video systems.

Synthetic Video Detector forms part of the NVIDIA AI for Media platform and can be deployed in on-premises, edge, hybrid and approved air-gapped environments.

NVIDIA said the flexible deployment options could help broadcasters, government agencies and other organisations analyse sensitive footage while retaining control over their data.

Why does it matter?

Synthetic video is becoming harder to identify during fast-moving news events, increasing pressure on media organisations to verify footage before publication. NVIDIA’s detector could provide an additional screening signal and help editorial teams focus limited verification resources. Still, its declining performance on compressed material shows why automated detection cannot replace source checks, forensic analysis and human judgement.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our chatbot!

Google launches Gemini 3.6 Flash for more efficient AI agents

Google has launched Gemini 3.6 Flash, describing it as a faster and more efficient model for developers building AI agents, coding tools and enterprise workflows while reducing the cost of deploying AI at scale.

According to Google, the model uses 17% fewer output tokens than Gemini 3.5 Flash on the Artificial Analysis Index and requires fewer reasoning steps and tool calls for multi-stage tasks, reducing the cost of running production AI agents.

Google has priced Gemini 3.6 Flash at US$1.50 per million input tokens and US$7.50 per million output tokens. The company says it improves coding, computer use, document analysis, chart interpretation and report drafting while reducing unnecessary edits and execution loops.

The company also reported gains on benchmarks including DeepSWE and OSWorld-Verified, alongside stronger safeguards against cyber-offence and chemical, biological, radiological and nuclear misuse.

Google also introduced Gemini 3.5 Flash-Lite, its fastest and lowest-cost model in the 3.5 family, targeting high-volume workloads such as search, document processing and data extraction.

In addition, a specialised Gemini 3.5 Flash Cyber model will power Google’s CodeMender security agent to detect, validate and patch software vulnerabilities. Owing to its potential for misuse, access will initially be restricted to governments and trusted partners through a pilot programme.

Gemini 3.6 Flash and Flash-Lite are available through the Gemini API, Google AI Studio, Android Studio, Gemini Enterprise and the Gemini app, with Flash-Lite also rolling out in Google Search.

Google said Gemini 3.5 Pro remains in partner testing while work is already underway on what it describes as its most ambitious pre-training run yet for Gemini 4.

Why does it matter?

The launch reflects a broader shift in AI development from simply increasing model size towards improving efficiency, reliability and cost-effectiveness for real-world deployment. As AI agents become more widely adopted, reducing inference costs and improving task execution are becoming increasingly important competitive advantages.

The restricted release of Gemini’s cybersecurity model also illustrates how AI developers are adopting more differentiated access policies for high-risk capabilities. Rather than making every model broadly available, companies are increasingly limiting advanced cyber tools to trusted users while attempting to balance defensive benefits with concerns about misuse.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Singapore expands AI governance training partnership with IAPP

Singapore’s Infocomm Media Development Authority (IMDA) and the International Association of Privacy Professionals (IAPP) have signed a three-year agreement to expand AI governance training, reflecting growing demand for professionals who can oversee the safe and responsible deployment of AI.

The Memorandum of Intent aims to equip regulators, industry leaders and practitioners with the skills needed to address increasingly complex issues related to AI governance, responsible data use and emerging technology regulation.

Under the partnership, professionals in Singapore will gain broader access to training and certification, including the IAPP’s AI Governance Professional programme. The organisations said demand is growing for specialists who can help organisations manage AI risks while supporting innovation.

IMDA and the IAPP will also continue integrating the Singapore Data Festival with the IAPP Asia Forum for another three years. The joint event will bring together policymakers, regulators and industry experts to discuss developments in AI governance, data protection and digital responsibility, reinforcing Singapore’s position as a regional hub for trusted digital governance.

The agreement was signed by Denise Wong, Commissioner of Singapore’s Personal Data Protection Commission and IMDA Assistant Chief Executive, and IAPP President and CEO J. Trevor Hughes.

Wong said the partnership would equip regulators and industry leaders with practical skills while attracting more international expertise to Singapore, while Hughes emphasised the growing importance of trained professionals as AI transforms industries.

The partnership will also promote discussions on ethical AI, trustworthy systems and organisational accountability, with both organisations aiming to help practitioners translate high-level governance principles into operational practice.

Why does it matter?

As AI regulation becomes more sophisticated, organisations increasingly need professionals who can translate legal requirements and ethical principles into practical governance processes. Training and certification programmes are becoming an important part of building the institutional capacity needed for responsible AI deployment.

The partnership also reinforces Singapore’s ambition to position itself as a regional centre for AI governance and digital trust. By combining international certification with local policy initiatives, it aims to strengthen the expertise needed to support AI adoption across both the public and private sectors.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

NVIDIA and Wistron open $700 million AI chip factory in Texas

Wistron has opened a $700 million AI manufacturing facility in Fort Worth, Texas, to produce advanced NVIDIA computing systems, strengthening domestic capacity for the AI infrastructure underpinning large-scale AI.

The 324,000-square-foot facility is Wistron’s first manufacturing facility in the United States. It currently operates one production line for NVIDIA’s GB300 Grace Blackwell Ultra Superchip, with a second line planned for the upcoming Vera Rubin platform.

Wistron expects output to reach tens of thousands of boards per month during 2026. More than 500 jobs have already been created, with the workforce expected to grow to 1,000 by the end of the year.

NVIDIA CEO Jensen Huang described domestic manufacturing as central to rebuilding US industrial capacity, noting the company’s commitment to producing up to US$500 billion worth of advanced AI platforms in the country.

The facility was designed and tested using a digital twin before construction began. Wistron employed NVIDIA’s Omniverse platform together with the Nemotron and Cosmos models, Metropolis libraries and PhysicsNeMo framework to simulate production lines, optimise factory layouts and train workers virtually, allowing engineers to validate assembly processes before physical equipment was installed.

During the opening ceremony, Wistron unveiled the first GB300 Grace Blackwell Ultra Superchip assembled at the site. According to Huang, systems based on the processor contain around 1.5 million components, weigh approximately two tonnes and cost about US$4 million.

The investment is intended to strengthen US supply chains and expand domestic capacity for producing AI infrastructure.

Why does it matter?

The new facility reflects a broader shift in AI policy from focusing primarily on software and models towards strengthening the industrial infrastructure needed to support them. As governments increasingly view AI hardware as a strategic asset, domestic manufacturing capacity is becoming an important component of economic security and technological competitiveness.

The project also illustrates how AI is transforming manufacturing itself. By using digital twins, simulation and AI models to design and optimise production before construction was completed, Wistron demonstrates how AI is reshaping factory operations while simultaneously producing the hardware that will power future AI systems.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Singapore expands AI strategy to accelerate enterprise adoption and workforce skills

Singapore is sharpening its AI strategy with a stronger emphasis on accelerating business adoption, expanding AI skills and strengthening governance, as the country seeks to translate AI capabilities into broader economic impact.

Speaking at IBM Think on Tour Singapore, Minister Josephine Teo said the updated approach builds on the National AI Strategy by narrowing the gap between widespread individual AI use and slower enterprise adoption.

Initiatives include the National AI Impact Programme, AI training for digital leaders, support for AI ‘bilinguals’ who combine domain expertise with AI knowledge, and expanded compute resources for organisations developing tailored AI applications.

Singapore is also working to make AI adoption more inclusive by helping professionals and SMEs access AI tools, skills and computing resources. Partnerships with technology companies will support organisations in identifying practical use cases and integrating AI into their operations.

Alongside expanding AI adoption, Singapore is strengthening governance frameworks to address risks associated with increasingly capable agentic AI systems. The country is also investing in quantum technologies as part of its broader ambition to position itself as a global AI innovation hub.

Why does it matter?

Singapore’s updated strategy reflects a broader shift in national AI policies from developing technical capabilities to accelerating adoption across the economy. Increasingly, governments are focusing not only on creating AI technologies but also on ensuring that businesses, workers and public institutions have the skills, infrastructure and support needed to use them effectively.

The strategy also demonstrates how AI competitiveness is becoming closely linked with governance and workforce development. By combining investment in AI adoption, talent and risk management, Singapore aims to strengthen its position as a regional AI hub while preparing for more advanced technologies such as agentic AI and quantum computing.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

India expands BHASHINI to support multilingual AI for digital public services

India is expanding its BHASHINI language AI platform to Goa, where it will support digital public services in Konkani and other Indian languages as part of the country’s broader multilingual digital public infrastructure strategy.

BHASHINI provides AI-powered translation, speech recognition, text-to-speech and transliteration, enabling government platforms and digital services to operate across multiple Indian languages. The platform supports applications in healthcare, education, tourism and citizen services, helping make public services more accessible across linguistic communities.

Discussions in Goa also focused on expanding Konkani language datasets to improve AI model performance and support the development of locally relevant digital applications. The initiative encourages collaboration between government agencies, researchers and technology developers to strengthen language resources for AI.

BHASHINI already supports dozens of Indian and international languages and has been integrated into hundreds of government websites, reflecting India’s broader strategy of using multilingual AI to strengthen digital public infrastructure and improve access to government services.

Why does it matter?

The expansion of BHASHINI illustrates how India is using AI to make digital public infrastructure more inclusive in a linguistically diverse country. By enabling government services to operate across multiple languages, the platform can help reduce language barriers that often limit access to healthcare, education and other public services.

The initiative also highlights the growing importance of language data as public digital infrastructure. Improving datasets for regional languages such as Konkani not only enhances AI performance but also helps ensure that smaller linguistic communities are represented as AI technologies become more widely integrated into public administration.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

EU releases multilingual AI model for all 24 official languages

The European Commission’s Directorate-General for Translation (DGT) has released an open large language model designed to provide balanced performance across all 24 official EU languages. Named the EU-Institutional-LLM, it is available in both base and instruct versions through the European Language Data Space.

Unlike many large language models that perform best in only a handful of languages, the EU-Institutional-LLM was designed to provide balanced support across all official EU languages while understanding the legislation, terminology and policy context specific to the Union. The model is trained using DGT’s high-quality multilingual datasets.

The base model is a continually pretrained version of Mixtral-8x7B-v0.1 with around 47 billion parameters and a Mixture-of-Experts architecture. It was trained to improve multilingual performance while limiting catastrophic forgetting through replay data. The instruct version was then created through supervised fine-tuning followed by preference alignment using odds ratio preference optimisation.

Training relied on European supercomputing infrastructure, including the Leonardo Booster system through the EuroHPC Joint Undertaking (EuroHPC JU), together with the MeluXina and MareNostrum supercomputers in Luxembourg, Italy and Spain.

Why does it matter?

The release addresses a longstanding challenge in multilingual AI. Most frontier language models are optimised for a relatively small number of widely spoken languages, creating uneven performance across multilingual societies. By designing a model specifically for all 24 official EU languages and embedding knowledge of EU legislation and institutions, the Commission is seeking to make AI more accessible and useful throughout the Union’s public sector.

The project also illustrates Europe’s broader strategy of building digital sovereignty through open AI infrastructure. Rather than relying exclusively on proprietary foreign models, the EU is investing in publicly available models, European datasets and domestic supercomputing resources that can support research, translation, public administration and future AI applications aligned with European values and regulatory frameworks.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

India expands Ayush Grid digital health infrastructure

India’s Ministry of Ayush has reported progress in expanding the Ayush Grid, a digital public infrastructure designed to support healthcare delivery, research, regulation, education and public services across the country’s traditional medicine sector.

The ministry said it has developed 24 e-governance platforms, including the My Ayush Integrated Services Portal, which brings together ministry services within a single digital ecosystem and features a beta AI chatbot to assist users.

The Ayush Hospital Management Information System supports patient registration, electronic health records, referrals and continuity of care. As of 15 July 2026, the platform had been adopted by 5,386 Ayush healthcare facilities.

Other digital platforms include the Ayush Research Portal, containing more than 43,000 curated publications, the Ayush Anudan funding portal, the Ayush Nivesh Saarthi investment gateway and the e-Charak medicinal plant marketplace.

Additional services support yoga programmes, health assessments and infrastructure planning. More than 760,000 organisations registered for Yoga Sangam during the International Day of Yoga 2026, while 6,002 facilities have been mapped through the PM GatiShakti Ayush Asset Mapping Tool.

The Swasthya Assessment Scale has also enrolled 1,335 doctors across 221 facilities.

The ministry plans to strengthen interoperability, multilingual services and data-driven governance while expanding the use of AI and machine learning. Future work will also focus on closer integration with national digital health initiatives, including the Ayushman Bharat Digital Mission, to improve access to Ayush services, particularly in rural and underserved areas.

Why does it matter?

The Ayush Grid demonstrates how India is extending its digital public infrastructure approach beyond conventional healthcare into traditional medicine. By integrating clinical services, research, funding and citizen-facing platforms, the initiative aims to improve coordination, transparency and continuity of care across the sector.

The planned expansion of interoperability, AI and multilingual digital services also reflects a broader shift towards data-driven health governance. Closer integration with national digital health programmes could improve access to healthcare while strengthening the role of traditional medicine within India’s wider digital health ecosystem.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

South Korea expands AI diplomacy through new development package

South Korea has unveiled the ‘K-AI Package’, a development cooperation strategy that combines AI infrastructure, digital capacity development and development finance to support AI adoption in developing countries while expanding international opportunities for Korean technology companies.

The strategy was presented on 20 July during a ministerial meeting on global economic affairs chaired by Deputy Prime Minister and Minister of Economy and Finance Koo Yun Cheol.

The initiative combines AI solutions, AI data centres and renewable energy facilities with infrastructure projects financed through the Economic Development Cooperation Fund. It covers seven sectors, including water management, healthcare, energy, transport, agriculture, culture and education, with projects ranging from AI-supported dam management and medical diagnosis systems to smart farming, renewable energy technologies and AI education programmes.

South Korea plans to integrate the package with existing development cooperation mechanisms, including the Knowledge Sharing Program, trust funds, development finance and joint financing with multilateral development banks.

Alongside infrastructure projects, the government will support partner countries in developing AI legislation, standards and institutions while investing in technical education, workforce training and local capacity development.

From the second half of 2026, South Korea will identify partner countries’ priorities and propose tailored K-AI projects before moving to implementation. AI will also become a priority under the Knowledge Sharing Program in 2027, while a proposed global AI hub will support technology adoption and skills development in emerging economies.

Why does it matter?

The K-AI Package illustrates how development cooperation is becoming an increasingly important instrument of AI diplomacy. Rather than focusing solely on technology exports, South Korea is combining infrastructure, policy advice, skills development and financing to help partner countries build AI capacity.

The initiative also reflects growing international competition to shape the global AI ecosystem through development partnerships. By linking development assistance with industrial strategy, South Korea aims to expand opportunities for its AI industry while strengthening long-term digital cooperation with emerging economies.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

ILO supports digital workplace safety tool in Malaysia

The International Labour Organization (ILO) has supported the launch of a digital occupational safety and health (OSH) assessment tool in Malaysia to help employers strengthen workplace safety, improve regulatory compliance and better protect workers.

Developed by the Malaysian Employers Federation (MEF) with support from the EU-funded PROTECT project, the tool was introduced during practical training for more than 80 employers in Port Klang.

The MEF–ILO OSH Digital Assessment Tool enables employers to conduct occupational safety and health assessments through a structured digital platform, while the accompanying OSHAWA Handbook provides practical guidance on implementing workplace safety measures.

Together, they aim to improve risk identification, strengthen compliance with Malaysian legislation and promote a stronger workplace safety culture across industries.

The initiative builds on Malaysia’s strengthened occupational safety framework following amendments to the Occupational Safety and Health Act 1994, which expanded workplace coverage and introduced more robust risk assessment requirements.

It also supports responsible business practices by helping employers meet growing international expectations on labour rights, human rights due diligence and workplace safety within global supply chains.

A particular focus of the project is improving protection for migrant workers, who make up a significant share of the workforce in sectors including manufacturing, construction, plantations and services.

By integrating the digital assessment tool into MEF’s long-term advisory and training services, the ILO aims to expand its use across Malaysia and support lasting improvements in workplace safety..

Why does it matter?

The initiative illustrates how digital tools are increasingly being used to strengthen labour governance by making workplace risk assessments more systematic, consistent and accessible. Digital platforms can also help employers monitor compliance more effectively and respond more quickly to occupational safety risks.

The project further reflects growing international expectations that businesses demonstrate responsible labour practices throughout global supply chains. By supporting compliance with both national legislation and international standards, digital workplace safety tools can improve worker protection while helping companies meet evolving regulatory and due diligence requirements.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!