Artificial intelligence

Character.AI and Google face suits over child safety claims

Three lawsuits have been filed in US federal courts alleging that Character.AI and its founders, with Google’s backing, deployed predatory chatbots that harmed children. The cases involve the family of 13-year-old Juliana Peralta, who died by suicide in 2023, and two other minors.

The complaints say the chatbots were designed to mimic humans, build dependency, and expose children to sexual content. Using emojis, typos, and pop-culture personas, the bots allegedly gained trust and encouraged isolation from family and friends.

Juliana’s parents say she engaged in explicit chats, disclosed suicidal thoughts, and received no intervention before her death. Nina, 15, from New York, attempted suicide after her mother blocked the app, while a Colorado, US girl known as T.S. was also affected.

Character.AI and Google are accused of misrepresenting the app as child-safe and failing to act on warning signs. The cases follow earlier lawsuits from the Social Media Victims Law Center over similar claims that the platform encouraged harm.

SMVLC founder Matthew Bergman stated that the cases underscore the urgent need for accountability in AI design and stronger safeguards to protect children. The legal team is seeking damages and stricter safety standards for chatbot platforms marketed to minors.

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

Researchers for OpenAI and Apollo find scheming behaviour in frontier AI models

OpenAI and Apollo Research developed evaluations to reveal hidden misalignment in advanced AI systems, often labelled scheming, and observed scheming-like behaviour in several frontier models, including OpenAI o3, o4-mini, Gemini 2.5 pro and Claude Opus-4.

Researchers treated covert actions, such as deliberately withholding or distorting task-relevant information, as a measurable proxy and stress-tested models across dozens of scenarios.

Scheming occurs when a model outwardly appears aligned while pursuing an alternate goal. The teams used a trader analogy to explain the risk: a rule-breaking broker who covers tracks can seem more compliant than an honest operator who accepts lower gains.

They trained models with deliberative alignment to counter the behaviour, asking them to read and reason about an anti-scheming specification before acting. Hence, models follow safety principles instead of merely avoiding detection.

That training produced about a thirtyfold reduction in covert actions on out-of-distribution tests, with o3 falling from 13 percent to 0.4 percent and o4-mini dropping from 8.7 percent to 0.3 percent.

Nevertheless, rare but serious failures remained, and measurement proved complex. Models often became more aware of being evaluated, and situational awareness can suppress observable scheming without altering underlying aims.

The study relied heavily on access to the models’ chain of thought to probe motivations, so preserving reasoning transparency is crucial for reliable oversight. The teams warned that training practices eroding transparency could make monitoring far harder and let misalignment hide rather than vanish.

OpenAI and Apollo called for broader cross-lab safety evaluations, stronger monitoring tools and continued research into anti-scheming techniques. They renewed their partnership, launched a $500,000 red-teaming challenge focused on scheming and proposed shared testing protocols.

The researchers emphasised there is no evidence that today’s deployed AI models would abruptly begin harmful scheming. Still, the risk will grow as systems take on more ambiguous, long-term, real-world responsibilities instead of short, narrow tasks.

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

Google adds AI features to Chrome browser on Android and desktop

Alphabet’s Google has announced new AI-powered features for its Chrome browser that aim to make web browsing more proactive instead of reactive. The update centres on integrating Gemini, Google’s AI assistant, into Chrome to provide contextual support across tabs and tasks.

The AI assistant will help students and professionals manage large numbers of open tabs by summarising articles, answering questions, and recalling previously visited pages. It will also connect with Google services such as Docs and Calendar, offering smoother workflows on desktop and mobile devices.

Chrome’s address bar, the omnibox, is being upgraded with AI Mode. Users can ask multi-part questions and receive context-aware suggestions relevant to the page they are viewing. Initially available in the US, the feature will roll out to other regions and languages soon.

Beyond productivity, Google is also applying AI to security and convenience. Chrome now blocks billions of spam notifications daily, fills in login details, and warns users about malicious apps.

Future updates are expected to bring agentic capabilities, enabling Chrome to carry out complex tasks such as ordering groceries with minimal user input.

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

Microsoft builds the world’s most powerful AI data centre in Wisconsin

US tech giant, Microsoft, is completing the construction of Fairwater in Mount Pleasant, Wisconsin, which it says will be the world’s most powerful AI data centre. The facility is expected to be operational in early 2026 after a $3.3 billion investment, with an additional $4 billion now committed for a second site.

The company says the project will help shape the next generation of AI by training frontier models with hundreds of thousands of NVIDIA GPUs, offering ten times the performance of today’s fastest supercomputers.

Beyond technology, Microsoft is highlighting the impact on local jobs and skills. Thousands of construction workers have been employed during the build, while the site is expected to support around 500 full-time roles when the first phase opens, rising to 800 once the second is complete.

The US giant has also launched Wisconsin’s first Datacentre Academy with Gateway Technical College to prepare students for careers in the digital economy.

Microsoft is also stressing its sustainability measures. The data centre will rely on a closed-loop liquid cooling system and outside air to minimise water use, while all fossil-fuel power consumed will be matched with carbon-free energy.

A new 250 MW solar farm is under construction in Portage County to support the commitment. The company has partnered with local organisations to restore prairie and wetland habitats, further embedding the project into the surrounding community.

Executives say the development represents more than just an investment in AI. It signals a long-term commitment to Wisconsin’s economy, education, and environment.

From broadband expansion to innovation labs, the company aims to ensure the benefits of AI extend to local businesses, students, and residents instead of remaining concentrated in global hubs.

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

Xbox app introduces Microsoft’s AI Copilot in beta

Microsoft has launched the beta version of Copilot for Gaming, an AI-powered assistant within the Xbox mobile app for iOS and Android. The early rollout covers over 50 regions, including India, the US, Japan, Australia, and Singapore.

Access is limited to users aged 18 and above, and the assistant currently supports English instead of other languages, with broader language support expected in future updates.

Copilot for Gaming is a second-screen companion, allowing players to stay informed and receive guidance without interrupting console gameplay.

The AI can track game activity, offer context-aware responses, suggest new games based on play history, check achievements, and manage account details such as Game Pass renewal and gamer score.

Users can ask questions like ‘What was my last achievement in God of War Ragnarok?’ or ‘Recommend an adventure game based on my preferences.’

Microsoft plans to expand Copilot for Gaming beyond chat-based support into a full AI gaming coach. Future updates could provide real-time gameplay advice, voice interaction, and direct console integration, allowing tasks such as downloading or installing games remotely instead of manually managing them.

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

WTO report notes AI’s potential benefit to trade if divides are addressed

The WTO launched the 2025 World Trade Report, titled ‘Making trade and AI work together to benefit all’. The report argues that AI could potentially boost global trade by up to 37% and GDP by 12–13% by 2040, particularly through digitally deliverable services.

It notes that AI can lower trade costs, improve supply-chain efficiency, and create opportunities for small firms and developing countries. Still, it warns that without deliberate action, AI could deepen global inequalities and widen the gap between advanced and developing economies.

The report underscores the need for investment in digital infrastructure, energy, skills, and enabling policies, highlighting the importance of IP protection, competition frameworks, and government support.

A newly developed indicator, the WTO AI Trade Policy Openness Index (AI-TPOI), revealed significant variation in AI-related trade policies across and within income groups.

It assessed three policy areas relevant to AI diffusion: barriers to services trade, restrictions on trade in AI-enabling goods, and limitations on cross-border data flows.

Stronger multilateral cooperation and targeted capacity-building were presented as essential to ensure AI-enabled trade supports inclusive, sustainable prosperity rather than reinforcing existing divides.

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

AI tool combines breast cancer and heart disease screening

Scientists from Australian universities and The George Institute for Global Health have developed an AI tool that analyses mammograms and a woman’s age to predict her risk of heart-related hospitalisation or death within 10 years.

Published in Heart on 17 September, the study highlights the lack of routine heart disease screening for women, despite cardiovascular conditions causing 35% of female deaths. The tool delivers a two-in-one health check by integrating heart risk prediction into breast cancer screening.

The model was trained on data from over 49,000 women and performs as accurately as traditional models that require blood pressure and cholesterol data. Researchers emphasise its low-resource nature, making it viable for broad deployment in rural or underserved areas.

Study co-author Dr Jennifer Barraclough said mobile mammography services could adopt the tool to deliver breast cancer and heart health screenings in one visit. Such integration could help overcome healthcare access barriers in remote regions.

Next, before a broader rollout, the researchers plan to validate the tool in more diverse populations and study practical challenges, such as technical requirements and regulatory approvals.

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

New Amazon AI transforms seller experience

Amazon has unveiled a significant upgrade to its Seller Assistant, evolving the tool into an agentic AI-powered partner that can actively help sellers manage and grow their businesses.

Powered by Amazon Bedrock and using advanced models from Amazon Nova and Anthropic Claude, the AI can respond to queries and plan, reason, and act with a seller’s permission. Independent sellers now have an assistant operating around the clock while controlling them.

The upgraded AI can optimise inventory, monitor account health, and provide strategic guidance on product listings and compliance requirements.

Analysing historical trends alongside current data can suggest new product categories, forecast demand, and propose advertising strategies to improve performance. Sellers can receive actionable recommendations instead of manually reviewing reports, saving time and effort.

Creative Studio also benefits from agentic AI capabilities, enabling sellers to generate professional-quality advertising content in hours instead of weeks.

The AI evaluates products alongside Amazon’s shopping signals and produces tailored ad concepts with clear reasoning, helping sellers refine campaigns and boost engagement. Early users report faster decisions, better inventory management, and more efficient marketing.

Amazon plans to extend Seller Assistant to other countries in the coming months at no extra cost.

The evolution highlights the growing role of AI in everyday business operations. It reflects Amazon’s commitment to integrating advanced technologies into the seller experience instead of relying solely on human intervention.

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

Huawei unveils roadmap for next-generation AI super pods

Huawei chairman Xu outlined the company’s roadmap for AI computing platforms, revealing plans to launch the Atlas 950 SuperPoD in Q4 2026. The system will use over 8,000 Ascend GPUs across 128 racks, covering 1,000 sq metres, and offer 6.7 times more computing power and 15 times more memory.

A year later, the Atlas 960 SuperPod will debut with up to 15,488 Ascend 960 chips, achieving 30 exaflops of computing power and 4,460TB of memory. Xu said the two systems will stay the world’s most potent super nodes, with uses beyond AI in general-purpose computing in China.

Huawei faces Western sanctions limiting access to advanced semiconductor nodes. Xu said assembling less advanced chips into super pods lets Huawei compete with rivals like Nvidia at a system level despite lower individual chip performance.

Over the next three years, Huawei will launch three new Ascend chip series: the 950 line, 950PR and 950DT, the 960, and the 970. The 950PR, optimised for early-stage inference and recommendations, will ship in Q1 2026, while the 950DT with 2Tb/s bandwidth will launch in Q4 2026.

The 960 will double its predecessor’s computing power and memory capacity and arrive in Q4 2027.

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

AI tool predicts risk of over 1,000 diseases years ahead

Scientists have unveiled an AI tool capable of predicting the risk of developing over 1,000 medical conditions. Published in Nature, the model can forecast certain cancers, heart attacks, and other diseases more than a decade in advance.

Developed by the German Cancer Research Centre (DKFZ), the European Molecular Biology Laboratory (EMBL), and the University of Copenhagen, the model utilises anonymised health data from the UK and Denmark. It tracks the order and timing of medical events to spot patterns that lead to serious illness.

Researchers said the tool is exceptionally accurate for diseases with consistent progression, including some cancers, diabetes, heart attacks, and septicaemia. Its predictions work like a weather forecast, indicating higher risk rather than certainty.

The model is less reliable for unpredictable conditions such as mental health disorders, infectious diseases, or pregnancy complications. It is more accurate for near-term forecasts than for those decades ahead.

Though not yet ready for clinical use, the system could help doctors identify high-risk patients earlier and enable more personalised, preventive healthcare strategies. Researchers say more work is needed to ensure the tool works for diverse populations.

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