Artificial intelligence

LLM shortcomings highlighted by Gary Marcus during industry debate

Gary Marcus argued at Axios’ AI+ Summit that large language models (LLMs) offer utility but fall short of the transformative claims made by their developers. He framed their fundamental role as groundwork for future artificial general intelligence. He suggested that meaningful capability shifts lie beyond today’s systems.

Marcus said alignment challenges stem from LLMs lacking robust world models and reliable constraints. He noted that models still hallucinate despite explicit instructions to avoid errors. He described current systems as an early rehearsal rather than a route to AGI.

Concerns raised included bias, misinformation, environmental impact and implications for education. Marcus also warned about the decline of online information quality as automated content spreads. He believes structural flaws make these issues persistent.

Industry momentum remains strong despite unresolved risks. Developers continue to push forward without clear explanations for model behaviour. Investment flows remain focused on the promise of AGI, despite timelines consistently shifting.

Strategic competition adds pressure, with the United States seeking to maintain an edge over China in advanced AI. Political signals reinforce the drive toward rapid development. Marcus argued that stronger frameworks are needed before systems scale further.

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

Google drives health innovation through new EU AI initiative

At the European Health Summit in Brussels, Google presented new research suggesting that AI could help Europe overcome rising healthcare pressures.

The report, prepared by Implement Consulting Group for Google, argues that scientific productivity is improving again, rather than continuing a long period of stagnation. Early results already show shorter waiting times in emergency departments, offering practitioners more space to focus on patient needs.

Momentum at the Summit increased as Google announced new support for AI adoption in frontline care.

Five million dollars from Google.org will fund Bayes Impact to launch an EU-wide initiative known as ‘Impulse Healthcare’. The programme will allow nurses, doctors and administrators to design and test their own AI tools through an open-source platform.

By placing development in the hands of practitioners, the project aims to expand ideas that help staff reclaim valuable time during periods of growing demand.

Successful tools developed at a local level will be scaled across the EU, providing a path to more efficient workflows and enhanced patient care.

Google views these efforts as part of a broader push to rebuild capacity in Europe’s health systems.

AI-assisted solutions may reduce administrative burdens, support strained workforces and guide decisions through faster, data-driven insights, strengthening everyday clinical practice.

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

Will the AI boom hold or collapse?

Global investment in AI has soared to unprecedented heights, yet the technology’s real-world adoption lags far behind the market’s feverish expectations. Despite trillions of dollars in valuations and a global AI market projected to reach nearly $5 trillion by 2033, mounting evidence suggests that companies struggle to translate AI pilots into meaningful results.

As Jovan Kurbalija argues in his recent analysis, hype has outpaced both technological limits and society’s ability to absorb rapid change, raising the question of whether the AI bubble is nearing a breaking point.

Kurbalija identifies several forces inflating the bubble, such as relentless media enthusiasm that fuels fear of missing out, diminishing returns on ever-larger computing power, and the inherent logical constraints of today’s large language models, which cannot simply be ‘scaled’ into human-level intelligence.

At the same time, organisations are slow to reorganise workflows, regulations, and skills around AI, resulting in high failure rates for corporate initiatives. A new competitive landscape, driven by ultra-low-cost open-source models such as China’s DeepSeek, further exposes the fragility of current proprietary spending and the vast discrepancies in development costs.

Looking forward, Kurbalija outlines possible futures ranging from a rational shift toward smaller, knowledge-centric AI systems to a world in which major AI firms become ‘too big to fail’, protected by government backstops similar to the 2008 financial crisis. Geopolitics may also justify massive public spending as the US and China frame AI leadership as a national security imperative.

Other scenarios include a consolidation of power among a handful of tech giants or a mild ‘AI winter’ in which investment cools and attention pivots to the next frontier technologies, such as quantum computing or immersive digital environments.

Regardless of which path emerges, the defining battle ahead will centre on the open-source versus proprietary AI debate. Both Washington and Beijing are increasingly embracing open models as strategic assets, potentially reshaping global standards and forcing big tech firms to rethink their closed ecosystems.

As Kurbalija concludes, the outcome will depend less on technical breakthroughs and more on societal choices, balancing openness, competition, and security in shaping whether AI becomes a sustainable foundation of economic life or the latest digital bubble to deflate under its own weight.

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

OpenAI launches nationwide AI initiative in Australia

OpenAI has launched OpenAI for Australia, a nationwide initiative to unlock the economic and societal benefits of AI. The program aims to support sovereign AI infrastructure, upskill Australians, and accelerate the country’s local AI ecosystem.

CEO Sam Altman highlighted Australia’s deep technical talent and strong institutions as key factors in becoming a global leader in AI.

A significant partnership with NEXTDC will see the development of a next-generation hyperscale AI campus and large GPU supercluster at Sydney’s Eastern Creek S7 site.

The project is expected to create thousands of jobs, boost local supplier opportunities, strengthen STEM and AI skills, and provide sovereign compute capacity for critical workloads.

OpenAI will also upskill more than 1.2 million Australians in collaboration with CommBank, Coles and Wesfarmers. OpenAI Academy will provide tailored modules to give workers and small business owners practical AI skills for confident daily use.

The nationwide rollout of courses is scheduled to begin in 2026.

OpenAI is launching its first Australian start-up program with local venture capital firms Blackbird, Square Peg, and AirTree to support home-grown innovation. Start-ups will receive API credits, mentorship, workshops, and access to Founder Day to accelerate product development and scale AI solutions locally.

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

EU partners with EIB to support AI gigafactories

The European Commission and the European Investment Bank Group (EIB) have signed a memorandum of understanding to support the development of AI Gigafactories across the EU. The partnership aims to position Europe as a leading AI hub by accelerating financing and the construction of large-scale AI facilities.

The agreement establishes a framework to guide consortia responding to the Commission’s informal Call for Expression of Interest. EIB advisory support will help turn proposals into bankable projects for the 2026 AI Gigafactory call, with possible co-financing.

The initiative builds on InvestAI, announced in February 2025, mobilising €20 billion to support up to five AI Gigafactories. These facilities will boost Europe’s computing infrastructure, reinforce technological sovereignty, and drive innovation across the continent.

By translating Europe’s AI ambitions into concrete, large-scale projects, the Commission and the EIB aim to position the EU as a global leader in next-generation AI, while fostering investment and industrial growth.

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

Google launches Workspace Studio for AI-powered automation

Google has made Workspace Studio generally available, allowing employees to design, manage, and share AI agents directly within Workspace. Powered by Gemini 3, these agents automate tasks ranging from simple routines to complex business workflows, all without coding.

The platform aims to save time on repetitive work, freeing employees to focus on higher-value activities.

Agents can understand context, reason through problems, and integrate with core Workspace apps such as Gmail, Drive, and Chat, as well as enterprise platforms like Asana, Jira, Mailchimp, and Salesforce.

Early adopters, including cleaning solutions leader Kärcher, have utilised Workspace Studio to streamline workflows, reducing planning time by up to 90% and consolidating multiple tasks into a single minute.

Workspace Studio allows users to build agents using templates or natural language prompts, making automation accessible to non-specialists. Agents can manage status reports, reminders, email triage, and critical tasks, such as legal notices or travel requests.

Teams can also easily share agents, ensuring collaboration and consistency across workflows.

The rollout to business customers will continue over the coming weeks. Users can start creating agents immediately, explore templates, use prompts for automations, and join the Gemini Alpha program to test early features and controls.

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

Campaigning in the age of generative AI

Generative AI is rapidly altering the political campaign landscape, argues the ORF article, which outlines how election teams worldwide are adopting AI tools for persuasion, outreach and content creation.

Campaigns can now generate customised messages for different voter groups, produce multilingual content at scale, and automate much of the traditional grunt work of campaigning.

On one hand, proponents say the technology makes campaigning more efficient and accessible, particularly in multilingual or resource-constrained settings. But the ease and speed with which content can be generated also lowers the barrier for misuse: AI-driven deepfakes, synthetic voices and disinformation campaigns can be deployed to mislead voters or distort public discourse.

Recent research supports these worries. For example, a large-scale study published in Science and Nature demonstrated that AI chatbots can influence voter opinions, swaying a non-trivial share of undecided voters toward a target candidate simply by presenting persuasive content.

Meanwhile, independent analyses show that during the 2024 US election campaign, a noticeable fraction of content on social media was AI-generated, sometimes used to spread misleading narratives or exaggerate support for certain candidates.

For democracy and governance, the shift poses thorny challenges. AI-driven campaigns risk eroding public trust, exacerbating polarisation and undermining electoral legitimacy. Regulators and policymakers now face pressure to devise new safeguards, such as transparency requirements around AI usage in political advertising, stronger fact-checking, and clearer accountability for misuse.

The ORF article argues these debates should start now, before AI becomes so entrenched that rollback is impossible.

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

New AI stroke-imaging tool halves time to treatment

A new AI-powered tool rolled out across England is helping clinicians diagnose strokes much sooner, significantly speeding up treatment decisions and improving patient outcomes. According to a study published in The Lancet Digital Health, roughly 15,000 patients benefited directly from AI-assisted scan reviews.

The tool, deployed at over 70 hospitals, analyses brain scans in minutes to rapidly identify clots, supporting doctors in deciding whether a patient needs urgent procedures such as a thrombectomy. Sites using the AI saw thrombectomy rates double (from 2.3% to 4.6%), compared with more modest increases at hospitals not using the technology.

Time is critical in stroke treatment: each 20-minute delay in thrombectomy reduces a patient’s chance of full recovery by around 1 per cent. The AI-driven system also helped cut the average ‘door-in to door-out’ time at primary stroke centres by 64 minutes, making it far more likely that patients reach a specialist centre in time for treatment.

Health-service leaders say the findings provide real-world evidence that AI imaging can save lives and reduce disability after stroke. As a result, the technology is now part of a wider national rollout across every regularly admitting stroke service in England.

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

Japanese high-schooler suspected of hacking net-cafe chain using AI

Authorities in Tokyo have issued an arrest warrant for a 17-year-old boy from Osaka on suspicion of orchestrating a large-scale cyberattack using artificial intelligence. The alleged target was the operator of the Kaikatsu Club internet-café chain (along with related fitness-gym business), which may have exposed the personal data of about 7.3 million customers.

According to investigators, the suspect used a computer programme, reportedly built with help from an AI chatbot, to send unauthorised commands around 7.24 million times to the company’s servers in order to extract membership information. The teenager was previously arrested in November in connection with a separate fraud case involving credit-card misuse.

Police have charged him under Japan’s law against unauthorised computer access and for obstructing business, though so far no evidence has emerged of misuse (for example, resale or public leaks) of the stolen data.

In his statement to investigators, the suspect reportedly said he carried out the hack simply because he found it fun to probe system vulnerabilities.

This case is the latest in a growing pattern of so-called AI-enabled cyber crimes in Japan, from fraudulent subscription schemes to ransomware generation. Experts warn that generative AI is lowering the barrier to entry for complex attacks, enabling individuals with limited technical training to carry out large-scale hacking or fraud.

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

Google boosts Nigeria’s AI development

The US tech giant, Google, has announced a $2.1 million Google.org commitment to support Nigeria’s AI-powered future, aiming to strengthen local talent and improve digital safety nationwide.

An initiative that supports Nigeria’s National AI Strategy and its ambition to create one million digital jobs, recognising the economic potential of AI, which could add $15 billion to the country’s economy by 2030.

The investment focuses on developing advanced AI skills among students and developers instead of limiting progress to short-term training schemes.

Google will fund programmes led by expert partners such as FATE Foundation, the African Institute for Mathematical Sciences, and the African Technology Forum.

Their work will introduce advanced AI curricula into universities and provide developers with structured, practical routes from training to building real-world products.

The commitment also expands digital safety initiatives so communities can participate securely in the digital economy.

Junior Achievement Africa will scale Google’s ‘Be Internet Awesome’ curriculum to help families understand safe online behaviour, while the CyberSafe Foundation will deliver cybersecurity training and technical assistance to public institutions, strengthening national digital resilience.

Google aims to create more opportunities similar to those of Nigerian learners who used digital skills to secure full-time careers instead of remaining excluded from the digital economy.

By combining advanced AI training with improved digital safety, the company intends to support inclusive growth and build long-term capacity across Nigeria.

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