Digital cooperation

Apple study finds AI fails on complex tasks

A recent study by Apple researchers exposed significant limitations in the capabilities of advanced AI systems and huge reasoning models (LRMs).

Apple’s team suggested this may point to a fundamental limit in how current AI models scale up to general reasoning.

These models, designed to solve complex problems through step-by-step thinking, experienced what the paper called a ‘complete accuracy collapse’ when faced with high-complexity tasks. Even when given an algorithm that should have ensured success, the models failed to deliver correct solutions.

The study found that LRMs performed well with low- and medium-difficulty tasks but deteriorated sharply as the complexity increased.

Rather than increasing their effort as problems became harder, the models reduced their reasoning paradoxically, leading to complete failure.

Experts, including AI researcher Gary Marcus and University of Surrey’s Andrew Rogoyski in the UK, called the findings alarming and indicative of a potential dead end in current AI development.

The study tested systems from OpenAI, Google, Anthropic and DeepSeek, raising serious questions about how close the industry is to achieving AGI.

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

OpenAI’s revenue almost doubles to $10 billion

OpenAI has revealed that its annualised revenue has surged to $10 billion as of June 2025, nearly doubling since December 2024, when it stood at $5.5 billion.

The rapid growth is driven by the widespread adoption of its ChatGPT AI models across consumer and business markets, putting the company on course to meet its earlier goal of $12.7 billion in revenue for the whole year.

The $10 billion figure excludes licensing income from Microsoft, a major investor, and some large one-off contracts, according to an OpenAI spokesperson. Despite recording a loss of about $5 billion last year, OpenAI’s impressive revenue scale places it well ahead of many rivals benefiting from the AI boom.

Other players in the AI space are also seeing strong growth. For instance, Anthropic recently surpassed $3 billion in annualised revenue, driven by startup demand using its code-generation models. Meanwhile, OpenAI plans to raise up to $40 billion in new funding, valuing the company at $300 billion.

Since launching ChatGPT over two years ago, OpenAI has expanded its offerings with various subscription plans and services. The company reported 500 million weekly active users as of March 2025, underscoring its dominant position in the AI market.

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

XRobotics makes 25,000 pizzas a month with AI

XRobotics, a San Francisco-based startup, is gaining traction in the food tech sector with its compact pizza-making robot, the xPizza Cube. Roughly the size of a washing machine, the machine uses AI to apply sauce, cheese and toppings, producing up to 100 pizzas per hour.

At $1,300 monthly on a three-year lease, it’s designed to fit seamlessly into existing kitchens and support, not replace, staff. Unlike failed predecessors such as Zume, XRobotics has found success by offering assistive rather than disruptive technology.

Their initial, larger model proved impractical, but the current compact version, launched in 2023, now produces 25,000 pizzas each month across an undisclosed number of customer locations. Both small pizzerias and large chains use the robot to cut labour time and improve consistency.

With over 73,000 pizza outlets in the US, XRobotics plans to stay focused on pizza for now. Its founders, self-professed pizza lovers, say customer response has confirmed that a well-targeted tool can significantly impact even the most traditional kitchens.

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

Apple reveals new AI features at WWDC

Apple has unveiled a range of AI features at its annual Worldwide Developers Conference, focusing on tighter privacy, enhanced user tools and broader integration with OpenAI’s ChatGPT. These updates will appear across iOS 26, iPadOS 26, macOS 26 and visionOS 26, set to launch in autumn.

While Apple Intelligence was first teased last year, the company now allows third-party developers to access its on-device AI models for the first time.

CEO Tim Cook and software chief Craig Federighi outlined how these features are intended to offer more personalised, efficient apps. Users of newer iPhones will benefit from tools such as live translation in Messages and FaceTime, and AI-powered image analysis via Visual Intelligence.

Apple also enables users to blend emojis creatively and use ChatGPT through its Image Playground to stylise photos. Enhancements to the Wallet app will help summarise order tracking from emails, and AI-generated voices will offer fitness updates.

Despite these innovations, Apple’s redesign of Siri remains incomplete and is not expected to launch soon.

The event failed to deliver major surprises, as many details had already been leaked. Investors responded cautiously, sending Apple shares down by 1.2%. The firm has lost 20% of its value in the year and no longer holds the top spot as the world’s most valuable company.

Nonetheless, Apple is expected to reveal more AI advancements in 2026.

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

China proposes rare earth export relief for EU

China has proposed creating a ‘green channel’ for rare earth exports to the EU, aiming to ease the impact of its recent restrictions. These materials, vital to electric vehicles and household appliances, have been under stricter export controls since April.

During recent talks, Trade Commissioner Maroš Šefčovič warned Chinese officials that the curbs had caused major disruptions across Europe, describing the situation as alarming. While some progress in licence approvals has been noted, businesses argue it remains inadequate.

The talks come as both sides prepare for a high-stakes EU-China summit and continue negotiations over tariffs on Chinese electric vehicles.

Brussels has imposed duties of up to 35.3%, citing unfair subsidies, while Beijing is pushing for a deal involving minimum pricing to avoid the tariffs.

China’s commerce ministry confirmed the discussions are in their final stage but acknowledged that more work is needed to reach a resolution.

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

UAE AI megaproject faces US chip export concerns

Plans for a vast AI data hub in the UAE have raised security concerns in Washington due to the country’s close ties with China.

The $100 billion Stargate UAE campus, aims to deploy advanced US chips, but US officials are scrutinising potential technology leakage risks.

Although the Trump administration supports the project, bipartisan fears remain about whether the UAE can safeguard US-developed AI and chips from foreign adversaries.

A final agreement has not been reached as both sides negotiate export conditions, with possible restrictions on Nvidia’s hardware.

The initial phase of the Stargate project will activate 200 megawatts of capacity by 2026, but the deal’s future may depend on the UAE’s willingness to accept strict US oversight.

Talks over potential amendments continue, delaying approval of what could become a $500 billion venture.

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

Meta plans $10 billion investment in Scale AI

Meta Platforms is reportedly in talks to invest over $10 billion in Scale AI, a data labelling startup already backed by Nvidia, Amazon, and Meta itself.

The deal, if finalised, would mark Meta’s largest external investment in AI to date, representing a notable shift away from its prior reliance on in-house research and open-source projects.

Founded in 2016, Scale AI supports the training of AI models through high-quality labelled datasets. It also provides a platform for AI research collaboration, now with contributors in more than 9,000 locations.

The company was last valued at nearly $14 billion following a 2024 funding round involving Meta and Microsoft.

Meta’s planned investment signals an aggressive expansion of its AI ambitions. Earlier this year, CEO Mark Zuckerberg announced up to $65 billion in AI spending for 2025. It includes Meta’s Llama chatbot, now embedded into Facebook, Instagram and WhatsApp, reaching one billion users monthly.

The move puts Meta in closer competition with Microsoft, which has committed over $13 billion to OpenAI, and Amazon and Alphabet, which are backing rival AI firm Anthropic. Scale AI declined to comment, while Meta has yet to respond publicly.

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

Anthropic debuts AI tools for US national security

Anthropic has launched a new line of AI models, Claude Gov, explicitly tailored for US national security operations. Built with direct input from government clients, top-tier agencies already use the models.

These classified-use models were developed with enhanced safety testing and are optimised for handling sensitive material, including improved handling of classified data, rare language proficiency, and defence-specific document comprehension.

The Claude Gov models reflect Anthropic’s broader move into government partnerships, building on its collaboration with Palantir and AWS.

As competition in defence-focused AI intensifies, rivals including OpenAI, Meta, and Google are also adapting their models for secure environments.

The sector’s growing interest in custom, security-conscious AI tools marks a shift in how leading labs seek stable revenue streams and deeper ties with government agencies.

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

New phase in Malaysia Airlines and Google collaboration

Malaysia Airlines is deepening its partnership with Google to accelerate its digital transformation, focusing on AI-powered marketing and end-to-end innovation across the travel experience.

Building on a successful 2024 collaboration, the renewed agreement will integrate more advanced technologies to enhance customer engagement and streamline services.

The airline and tech giant aims to shift from promotional showcases to a fully integrated ecosystem, using AI and data-driven tools to reshape the travel journey.

A recent highlight was an interactive showcase at the MATTA Fair, where visitors transformed photos into videos using Google’s Veo image-to-video generation tool — part of Malaysia Airlines’ ‘Time For’ campaign.

Dersenish Aresandiran, chief commercial officer of airlines at Malaysia Aviation Group, said the partnership is about more than technology: it’s a strategic move to enrich customer experiences and support national tourism goals ahead of Visit Malaysia Year 2026.

Google Malaysia’s director, Farhan Qureshi, echoed the sentiment, calling the collaboration a powerful example of how innovation and creativity can redefine global travel.

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

Epic adds AI NPC tools to Fortnite as Vader voice sparks union clash

Epic Games is launching new tools for Fortnite creators that enable them to build AI-powered non-player characters (NPCs), following the debut of an AI-generated Darth Vader that players can talk to in-game.

The feature, which reproduces the iconic voice of James Earl Jones using AI, marks a significant step in interactive gaming—but also comes with its share of challenges and controversy.

According to The Verge, Epic encountered several difficulties in fine-tuning Vader’s voice and responses to feel authentic and fit smoothly into gameplay. ‘The culmination of a very intense effort for a character everybody understands,’ said Saxs Persson, executive vice president of the Fortnite ecosystem.

Persson noted that the team worked carefully to ensure that when Vader joins a player’s team, he behaves as a fearsome and aggressive ally—true to his cinematic persona.

However, the rollout wasn’t entirely smooth. In a live-streamed session, popular Fortnite creator Loserfruit prompted Vader to swear, exposing the system’s content filtering flaws. Epic responded quickly with patches and has since implemented multiple layers of safety checks.

‘We do our best job on day one,’ said Persson, ‘but more importantly, we’re ready to surround the problem and have fixes in place as fast as possible.’

Now, Fortnite creators will have access to the same suite of AI tools and safety systems used to develop Vader. They can control voice tone, dialogue, and NPC behaviour while relying on Epic’s safeguards to avoid inappropriate interactions.

The feature launch comes at a sensitive moment, as actor union SAG-AFTRA has filed a complaint against Epic Games over using AI to recreate Vader’s voice.

The union claims that Llama Productions, an Epic subsidiary, employed the technology without consulting or bargaining with the union, replacing the work of human voice actors.

‘We must protect our right to bargain terms and conditions around uses of voice that replace the work of our members,’ SAG-AFTRA said, emphasising its support for actors and estates in managing the use of digital replicas.

As Epic expands its AI capabilities in gaming, it faces both the technical challenges of responsible deployment and the growing debate around AI’s impact on creative professions.

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