Access

China issues action plan for global AI governance and proposes global AI cooperation organisation

At the 2025 World AI Conference in Shanghai, Chinese Premier Li Qiang urged the international community to prioritise joint efforts in governing AI, making reference to a need to establish a global framework and set of rules widely accepted by the global community. He unveiled a proposal by the Chinese government to create a global AI cooperation organisation to foster international collaboration, innovation, and inclusivity in AI across nations.

‘China attaches great importance to global AI governance, and has been actively promoting multilateral and bilateral cooperation with a willingness to offer more Chinese solutions‘.

An Action Plan for AI Global Governance was also presented at the conference. The plan outlines, in its introduction, a call for ‘all stakeholders to take concrete and effective actions based on the principles of serving the public good, respecting sovereignty, development orientation, safety and controllability, equity and inclusiveness, and openness and cooperation, to jointly advance the global development and governance of AI’.

The document includes 13 points related to key areas of international AI cooperation, including promoting inclusive infrastructure development, fostering open innovation ecosystems, ensuring high-quality data supply, and advancing sustainability through green AI practices. It also calls for consensus-building around technical standards, advancing international cooperation on AI safety governance, and supporting countries – especially those in the Global South – in ‘developing AI technologies and services suited to their national conditions’.

Notably, the plan indicates China’s support for multilateralism when it comes to the governance of AI, calling for an active implementation of commitments made by UN member states in the Pact for the Future and the Global Digital Compact, and expressing support for the establishment of the International AI Scientific Panel and a Global Dialogue on AI Governance (whose terms of reference are currently negotiated by UN member states in New York).

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

US senator urges Musk to block Starlink use by Southeast Asian criminal networks

US Senator Maggie Hassan has called on SpaceX CEO Elon Musk to take immediate action against transnational criminal groups in Southeast Asia, which are allegedly using Starlink satellite internet to perpetrate massive online fraud schemes targeting American citizens.

In a letter seen by Reuters, the senator highlighted the growing role of Starlink in enabling so-called ‘scam compounds’ operated by criminal syndicates across Myanmar, Thailand, Cambodia, and Laos.

According to the US Treasury’s Financial Crimes Enforcement Network, the fraud networks have collectively cost Americans billions of dollars.

Senator Hassan emphasised that although SpaceX’s service rules allow for termination of access in cases of fraudulent activity, Starlink appears to remain active in regions where these scams flourish. She urged Musk to uphold SpaceX’s stated standards and take responsibility for cutting off illicit use of the service.

The scam compounds in question are more than just virtual hubs; reportedly, they are the sites of forced labour and human trafficking. Reports, including those from the UN, detail how hundreds of thousands of people have been trafficked into these centres, where they are coerced into operating elaborate online fraud schemes. These often target victims in the US and around the world through phishing messages, fake investment offers, and digital extortion.

The region has taken some steps to curb these operations. Since February, Thailand has actively disrupted resources such as electricity and internet to areas along its border with Myanmar, notably Myawaddy, where many scam centres are based. However, satellite services like Starlink can bypass these traditional infrastructure shutdowns, enabling fraud operations to persist despite regional crackdowns.

The criminal networks, many of which have roots in China, have also captured international attention due to high-profile cases. One such case was the January abduction of Chinese actor Wang Xing, who was kidnapped after arriving in Thailand and later rescued across the border in Myanmar by Thai authorities.

The incident further exposed these networks’ dangerous and organised nature, prompting broader calls for transnational cooperation and tech-sector accountability.

Source: Reuters

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

AI fuels new wave of global security breaches

Global corporations are under growing threat from increasingly sophisticated cyber attacks as AI tools boost the capabilities of malicious actors.

Allianz Life recently confirmed a breach affecting most of its 1.4 million North American customers, adding to a string of high-profile incidents this year.

Microsoft is also contending with the aftermath of a wide-scale intrusion, as attackers continue to exploit AI-driven methods to bypass traditional defences.

Cybersecurity firm DeepStrike reports that over 560,000 new malware samples are detected daily, underscoring the scale of the threat.

Each month in 2025 has brought fresh incidents. January saw breaches at the UN and Hewlett-Packard, while crypto lender zkLend lost $9.5 million to hackers in February.

March was marked by a significant attack on Elon Musk’s X platform, and Oracle lost six million data records.

April and May were particularly damaging for retailers and financial services. M&S, Harrods, and Coinbase were among the prominent names hit, with the latter facing a $20 million ransom demand. In June, luxury brands and media companies, including Cartier and the Washington Post, were also targeted.

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

AI startup Daydream revolutionises online fashion search

Online shopping for specific items like bridesmaid dresses can be challenging due to overwhelming choices. A new tech startup, Daydream, aims to simplify this. It uses AI to let users search for products by describing them in natural language, making the process easier and more intuitive.

For instance, a user could ask for a ‘revenge dress to wear to a party in Sicily in July,’ or ‘a summer bag to carry to work and cocktails after.’

Daydream, with staff based in New York and San Francisco, represents the latest venture in a growing trend of tech companies utilising AI to streamline and personalise online retail.

Consumer demand for such tools is evident: an Adobe Analytics survey of 5,000 US consumers revealed that 39% had used a generative AI tool for online shopping last year, with 53% planning to do so this year. Daydream faces competition from tech giants already active in this space.

Meta employs AI to facilitate seller listings and to target users with more relevant product advertisements. OpenAI has launched an AI agent capable of shopping across the web for users, and Amazon is trialling a similar feature.

Google has also introduced various AI shopping tools, including automated price tracking, a ‘circle to search’ function for identifying products in photos, and virtual try-on options for clothing.

Despite the formidable competition, Daydream’s CEO, Julie Bornstein, believes her company possesses a deeper understanding of the fashion and retail industries.

Bornstein’s extensive background includes helping build Nordstrom’s website as its vice president of e-commerce in the early 2000s and holding C-suite positions at Sephora and Stitch Fix. In 2018, she co-founded her first AI-powered shopping startup, The Yes, which was sold to Pinterest in 2022.

Bornstein asserts, ‘They don’t have the people, the mindset, the passion to do what needs to be done to make a category like fashion work for AI recommendations.’ She added, ‘Because I’ve been in this space my whole career, I know that having the catalogue with everything and being able to show the right person the right stuff makes shopping easier.’

Daydream has already secured $50 million in its initial funding round, attracting investors such as Google Ventures and model Karlie Kloss, founder of Kode With Klossy. The platform operates as a free, digital personal stylist.

Users can input their desired products using natural language, eliminating the need for complex Boolean search terms, thanks to its AI text recognition technology, or upload an inspiration photo.

Daydream then presents recommendations from over 8,000 brand partners, ranging from budget-friendly Uniqlo to luxury brand Gucci. Users can further refine their search through a chat interface, for example, by requesting more casual or less expensive alternatives.

As users interact more with the platform, it progressively tailors recommendations based on their search history, clicks, and saved items.

When customers are ready to purchase, they are redirected to the respective brand’s website to complete the transaction, with Daydream receiving a 20% commission on the sale.

Unlike many other major e-commerce players, Bornstein is deliberately avoiding ad-based rankings. She aims for products to appear on recommendation pages purely because they are a suitable match for the customer, not due to paid placements.

Bornstein stated, ‘As soon as Amazon started doing paid sponsorships, I’m like, ‘How can I find the real good product?’ She emphasised, ‘We want this to be a thing where we get paid when we show the customer the right thing.’

A recent CNN test of Daydream yielded mixed results. A search for a ‘white, fitted button-up shirt for the office with no pockets’ successfully returned a $145 cotton long-sleeve shirt from Theory that perfectly matched the description.

However, recommendations are not always flawless. A query for a ‘mother of the bride dress for a summer wedding in California’ presented several slinky slip dresses, some in white, alongside more formal styles, appearing more suitable for a bachelorette party.

Bornstein confirmed that the company continuously refined its AI models and gathered user feedback. She noted, ‘We want data on what people are doing so we can focus and learn where we do well and where we don’t.’

Part of this ongoing development involves training the AI to understand nuanced contextual cues, such as the implications of a ‘dress for a trip to Greece in August’ (suggesting hot weather) or an outfit for a ‘black-tie wedding’ (implying formality).

Daydream’s web version launched publicly last month, and it is currently in beta testing, with plans for an app release in the autumn. Bornstein envisions a future where AI extends beyond shopping, assisting with broader fashion needs like pairing new purchases with existing wardrobe items.

She concluded, ‘This was one of my earliest ideas, but I didn’t know the term (generative AI) and I didn’t know a large language model would be the unlock.’

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

AI chatbot captures veteran workers’ knowledge to support UK care teams

Peterborough City Council has turned the knowledge of veteran therapy practitioner Geraldine Jinks into an AI chatbot to support adult social care workers.

After 35 years of experience, colleagues frequently approached Jinks seeking advice, leading to time pressures despite her willingness to help.

In response, the council developed a digital assistant called Hey Geraldine, built on the My AskAI platform, which mimics her direct and friendly communication style to provide instant support to staff.

Developed in 2023, the chatbot offers practical answers to everyday care-related questions, such as how to support patients with memory issues or discharge planning. Jinks collaborated with the tech team to train the AI, writing all the responses herself to ensure consistency and clarity.

Thanks to its natural tone and humanlike advice, some colleagues even mistook the chatbot for the honest Geraldine.

The council hopes Hey Geraldine will reduce hospital discharge delays and improve patient access to assistive technology. Councillor Shabina Qayyum, who also works as a GP, said the tool empowers staff to help patients regain independence instead of facing unnecessary delays.

The chatbot is seen as preserving valuable institutional knowledge while improving frontline efficiency.

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

Z.ai unveils cheaper, advanced AI model GLM-4.5

Chinese AI startup Z.ai, formerly Zhipu, is increasing pressure on global competitors with its latest model, GLM-4.5. The company has adopted an aggressive open-source strategy to attract developers. Anyone can download and use the model without licensing fees or platform restrictions.

GLM-4.5 is designed with agentic AI, breaking tasks into smaller components for improved performance. By approaching problems step by step, the model delivers more accurate and efficient outcomes. Z.ai aims to stand out through both technical sophistication and affordability.

CEO Zhang Peng says the model runs on only eight Nvidia H20 chips, while DeepSeek’s model needs sixteen. Nvidia developed the H20 to comply with US export controls aimed at China. Reducing chip demand significantly lowers the model’s operational footprint.

Zhang said the company has enough computing power and is not seeking further hardware now. Z.ai plans to charge 11 cents per million input tokens, undercutting DeepSeek R1’s 14 cents. Output tokens will cost 28 cents per million, compared to DeepSeek’s 2.19 dollars.

Such pricing could reshape large language model deployment expectations, especially in resource-limited environments. High costs have long been a barrier to broader AI adoption. Z.ai appears to be positioning itself as a more accessible alternative.

Founded in 2019, Z.ai has raised more than 1.5 billion dollars from investors including Alibaba, Tencent, and Qiming Venture Partners. It has grown quickly from a research-focused lab to one of China’s most prominent AI contenders. A public listing in Greater China is reportedly being prepared.

OpenAI recently named Zhipu among the Chinese firms it considers strategically significant in global AI development. US authorities responded by restricting American companies from working with Z.ai. The startup has nonetheless continued to expand its model lineup and partnerships.

Chinese firms increasingly invest in open-source models, often with domestic hardware compatibility in mind. Moonshot, another Alibaba-backed company, released the Kimi K2 model. Kimi K2 has received praise for its performance in coding and mathematical tasks.

Tencent has joined the race with its HunyuanWorld-1.0 model, which is built to generate immersive 3D environments. The HunyuanWorld-1.0 can accelerate game development, virtual reality design, and simulation work. Cutting-edge features are being paired with highly efficient architectures.

Alibaba also introduced its Qwen3-Coder model to assist in code generation and debugging. Such AI tools are seeing increasing use in software engineering and education. Chinese developers are positioning themselves to compete with Western offerings such as OpenAI’s Codex and Anthropic’s Claude.

The momentum within China’s AI sector is accelerating despite geopolitical and trade restrictions. A clear shift is underway from imitation to innovation, with local startups advancing independent research. Many models are trained on China-specific datasets to optimise relevance and performance.

Z.ai’s strategy combines cost reduction, efficient chip use, and broad availability. The company can build community trust and encourage ecosystem growth by open-sourcing its tools. At the same time, pricing undercuts major rivals and could disrupt the market.

Global AI development is increasingly decentralised, with Chinese firms no longer just playing catch-up. Large-scale funding and state support are helping to close gaps in hardware and training infrastructure. Z.ai is one of several firms pushing toward greater technological autonomy.

Open-source AI development is also helping Chinese companies win favour with developers outside their borders. Many international teams are experimenting with Chinese models to diversify risk and reduce reliance on US tech. Z.ai’s GLM-4.5 is among the models gaining traction globally.

By offering a powerful, lightweight, and affordable model, Z.ai is setting a new benchmark in the industry. The combination of technical refinement and strategic pricing draws attention from investors and users. A new era of AI competition is emerging.

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

Allianz breach affects most US customers

Allianz Life has confirmed a major cyber breach that exposed sensitive data from most of its 1.4 million customers in North America.

The attack was traced back to 16 July, when a threat actor accessed a third-party cloud system using social engineering tactics.

The cybersecurity breach affected a customer relationship management platform but did not compromise the company’s core network or policy systems.

Allianz Life acted swiftly by notifying the FBI and other regulators, including the attorney general’s office in Maine.

Those impacted are offered two years of credit monitoring and identity theft protection. The company has begun contacting affected individuals but declined to reveal the full number involved due to an ongoing investigation.

No other Allianz subsidiaries were affected by the breach. Allianz Life employs around 2,000 staff in the US and remains a key player within the global insurer’s North American operations.

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

Huawei challenges Nvidia with AI super server

Huawei has unveiled its most powerful AI server, the CloudMatrix 384, to challenge Nvidia’s grip on the high-performance AI infrastructure market.

The system, launched at the World AI Conference in Shanghai, uses 384 Ascend 910C chips, significantly outnumbering Nvidia’s 72 B200 GPUs in the GB200 NVL72.

Although Nvidia’s GPUs remain more powerful individually, Huawei’s design relies on stacking and high-speed chip interconnection to boost overall performance.

The company claims the CloudMatrix 384 can deliver 300 petaflops of computing power, well above Nvidia’s 180 petaflops, though it consumes nearly four times more energy.

The US recently reversed its ban on Nvidia’s H20 chip exports to China, seeking to curb Huawei’s momentum. However, ongoing reports of smuggled Nvidia GPUs raise doubts over the effectiveness of these restrictions.

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

New AI model, Aeneas, assists historians in interpreting Roman inscriptions

Thanks to AI, historians studying ancient Rome now have a powerful new tool.

A research team, including scholars from Google DeepMind and the University of Nottingham, developed a generative AI model called Aeneas that can help interpret damaged Latin inscriptions by estimating their location and date and suggesting likely missing text.

Each year, roughly 1,500 new Latin inscriptions are unearthed, ranging from imperial decrees to everyday graffiti. These inscriptions, written by ancient Romans across all social classes, offer rare, first-hand insights into daily life, language, and society.

Yet many of them are incomplete or difficult to contextualise. Traditionally, scholars must compare each inscription against hundreds of others manually — a process described as laborious and requiring exceptional expertise.

Aeneas, trained on over 170,000 Latin texts, can now predict when and where an inscription was written across the Roman Empire’s 62 provinces. In one test case, it analysed the famous Res Gestae Divi Augusti, narrowing down the date to the same two options long debated by historians.

Aeneas significantly improved research outcomes when used alongside human expertise instead of replacing it, helping scholars piece together history more efficiently than ever.

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

Viasat launches global IoT satellite service

Viasat has unveiled a new global connectivity service designed to improve satellite-powered internet of things (IoT) communication, even in remote environments. The new offering, IoT Nano, supports industries like agriculture, mining, transport with reliable, low-data and low-power two-way messaging.

The service builds on Orbcomm’s upgraded OGx platform, delivering faster message speeds, greater data capacity and improved energy efficiency. It maintains compatibility with older systems while allowing for advanced use cases through larger messages and reduced power needs.

Executives at Viasat and Orbcomm believe IoT Nano opens up new opportunities by combining wider satellite coverage with smarter, more frequent data delivery. The service is part of Viasat’s broader effort to expand its scalable and energy-efficient satellite IoT portfolio.

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