Artificial intelligence Archives | Page 5 of 678

IWF backs Pope Leo XIV call for responsible AI development

The Internet Watch Foundation has welcomed Pope Leo XIV’s reflections on AI, arguing that AI systems must be developed with stronger safeguards to protect children from abuse.

In a blog post, the IWF said the Pope’s message that technology should serve the common good and remain subject to human judgement and accountability reflects the risks its analysts are already seeing online.

The organisation warned that AI is being used to generate highly realistic child sexual abuse images and videos at scale. It said the number of AI-generated child sexual abuse videos identified by the IWF in 2025 increased by more than 260%, with nearly two-thirds falling into the most severe category of abuse.

The IWF also raised concerns about AI-nudification tools, which can generate realistic sexualised images of children and other individuals. Following the Child Dignity in the Artificial Intelligence Era conference in Rome, the organisation joined more than 100 organisations and individuals in supporting calls for a global ban on such tools.

The IWF said AI safety should be built into products from the earliest stages of development. Through its Safety by Design work, the organisation is calling for companies to assess, test and mitigate risks before AI systems reach the public.

It also called for stronger regulation, global alignment and enforceable safety-by-design standards to prevent the creation and spread of AI-generated child sexual abuse material.

Why does it matter?

The IWF’s warning shows how generative AI is creating urgent child protection risks, especially through realistic synthetic abuse material and nudification tools. The issue is no longer only content moderation after harm occurs; it increasingly concerns model design, testing, deployment and accountability before AI systems reach users. That makes safety by design, developer responsibility and international coordination central to AI governance.

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

EPO and Jio discuss AI, 5G and 6G patent strategy

The European Patent Office held a high-level online meeting with Jio Platforms to discuss patent quality, innovation and developments in the European patent system.

The meeting formed part of the EPO’s 2026 Quality Action Plan and wider user engagement with companies using the European patent system. Jio Platforms, India’s largest mobile network operator, is described by the EPO as an increasingly active user of the system.

Discussions covered AI, 5G, future 6G technologies, satellite communications and automation. Jio highlighted Europe’s growing importance for its patent strategy and outlined a global patent portfolio covering India, Europe, the United States and other jurisdictions.

A significant part of the exchange focused on standards and standard-essential patents. The EPO said the discussion explored developments in 5G and future 6G standards, as well as the Office’s recent work and research in the field.

The meeting also covered the Unitary Patent and Unified Patent Court. Jio does not currently use the Unitary Patent system, but the company expressed interest in assessing it as a possible option for future European patent protection.

Why does it matter?

The meeting highlights how patent strategy is becoming part of global competition in AI, 5G, 6G and digital infrastructure. For companies active in next-generation networks, standard-essential patents can shape market access, licensing power and participation in technical standards. Jio’s engagement with the EPO also reflects the growing internationalisation of Indian technology firms as they seek protection for innovation in European markets.

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

US backs photonics expansion for AI data centres under CHIPS Act

The Department of Commerce’s CHIPS Program Office has signed a letter of intent to provide up to $50 million in direct funding to Coherent Corp. under the CHIPS and Science Act.

According to the CHIPS Program Office, the proposed funding would support the expansion of Coherent’s facility in Sherman, Texas, which it describes as the first and largest high-volume 150mm indium phosphide semiconductor manufacturing facility in the United States.

The expansion would add wafer fabrication equipment and cleanroom capacity to increase production of indium phosphide-based photonic components. These components are used in high-speed optical interconnects that enable rapid data transfer within advanced AI data centres.

The Department of Commerce said the project would create high-skilled manufacturing jobs and strengthen domestic supply chains for critical photonics technologies that support next-generation computing and AI infrastructure.

Why does it matter?

The announcement highlights the growing importance of photonics technologies in the AI economy. As demand for AI computing continues to rise, data centres require increasingly efficient methods for transferring vast amounts of information between processors, servers and storage systems. Optical interconnect technologies based on indium phosphide semiconductors are becoming a critical part of that infrastructure.

The investment also reflects broader US industrial policy goals under the CHIPS and Science Act. Beyond traditional semiconductor manufacturing, policymakers are increasingly targeting specialised components and supply chains considered strategically important for AI competitiveness, economic security and technological resilience.

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

Japan and the Philippines partner on an AI-powered disaster risk platform

The Japan International Cooperation Agency (JICA) and the Philippines Department of Science and Technology (DOST) have agreed to collaborate on the GATES programme, a national geospatial and AI initiative aimed at strengthening disaster risk reduction and data-driven governance.

The Geospatial Analytics & Technology Solutions (GATES) programme is led by DOST and aims to integrate fragmented disaster-risk, geospatial, climate, hazard and strategic datasets from across government into a unified and interoperable data ecosystem. The programme uses geospatial analytics and artificial intelligence to support digital transformation, evidence-based policymaking, and science, technology, and innovation.

The collaboration will be carried out through the JICA Digital Transformation Lab and will begin in June 2026 for around four months. JICA DXLab will provide technical expertise to support data interoperability, governance frameworks and digital transformation capabilities under the GATES programme.

The partnership will also explore how the platform could contribute to broader digital public infrastructure for climate and disaster-risk information. DOST and JICA said the goal is to build a secure, reliable, and trusted AI ecosystem for disaster risk reduction and data-driven governance.

The collaboration builds on a long-running Japan–Philippines partnership in disaster resilience. Since the 1970s, JICA has supported the Philippine Atmospheric, Geophysical and Astronomical Services Administration through meteorological equipment and technical cooperation, helping strengthen observation, forecasting, and early warning systems for typhoons, heavy rainfall, and flooding.

JICA has also worked with the Philippine Institute of Volcanology and Seismology, including through the 2004 Metro Manila Earthquake Impact Reduction Study and ongoing cooperation on seismic monitoring, early warning systems, and data analysis capabilities.

Launched in 2025 as part of DOST’s ELEV8PH initiative, GATES is a three-year programme designed to create an AI-ready national disaster-risk data platform to support decision-making. The programme has already deployed the HANDA integrated preparedness platform, developed an initial data architecture blueprint, and started capacity-building and training activities.

The engagement has two main objectives. The first is to strengthen the platform’s data foundations by helping DOST connect disaster-risk datasets across agencies while preserving agency ownership, developing standards for AI-ready data and establishing a framework for responsible AI use.

The second objective is to turn disaster risk data into decisions. JICA will work with DOST and partner agencies to identify use cases that can improve disaster response and climate risk management, including tools for modelling incoming typhoons, dashboards for tracking climate-related risks, and planning aids for evacuation.

The cooperation is expected to strengthen disaster-risk management systems, support data-driven policymaking and expand the use of disaster-risk data for resilience and sustainable development. The partners also expect the work to strengthen the use of data and AI in policymaking in both the Philippines and Japan.

Why does it matter?

The initiative highlights the growing role of AI, geospatial analytics and interoperable data systems in disaster resilience and climate adaptation. Countries vulnerable to extreme weather events increasingly rely on integrated data platforms to improve forecasting, early warning systems, evacuation planning and emergency response.

The partnership also demonstrates how digital public infrastructure is evolving beyond identity and payments to include critical public-interest datasets. By creating an AI-ready and interoperable disaster-risk data ecosystem, the Philippines could strengthen evidence-based policymaking, improve coordination across government agencies and provide a model for other climate-vulnerable countries seeking to use AI and data more effectively in public administration.

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

UK unveils AI tools to speed up planning decisions and housing delivery

The Department for Science, Innovation and Technology (DSIT) and the Ministry of Housing, Communities and Local Government have unveiled two AI tools designed to modernise England’s planning system and accelerate housing delivery.

One new AI prototype is being tested by Barnet, Camden and Dorset councils and aims to reduce average decision times for routine householder planning applications from eight weeks to four. The system triages applications and provides planning officers with preliminary assessments to support decision-making.

A second tool, called Extract, has been made available to local authorities across England. It uses AI to convert decades of planning documents and maps into structured digital data, reducing the need for manual processing and allowing staff to focus on more complex cases.

The government said the initiatives support its target of building 1.5 million homes during this Parliament while improving the efficiency of public services through technology. Subject to successful trials, the new planning application tool is expected to be rolled out nationally in England from 2027.

Why does it matter?

The initiative illustrates how governments are increasingly using AI to address administrative bottlenecks and improve public-service delivery. Planning systems often face challenges related to outdated records, resource constraints and lengthy approval processes, making them a key target for digital transformation efforts.

The UK’s approach also highlights the growing role of AI in housing and infrastructure policy. If successful, the tools could help accelerate housing development, improve the use of public-sector resources and demonstrate how AI can support decision-making while leaving final judgments in the hands of public officials.

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

New AI breakthrough in cardiology balances patient data privacy and diagnosis

Researchers at the University of Kansas have developed a new AI model designed to improve the analysis of electrocardiogram (ECG) data while strengthening protections for patient privacy. The innovation responds to growing concerns that AI-enhanced ECGs can reveal sensitive personal attributes beyond heart activity.

The model, known as PP-VAE, aims to preserve clinically relevant insights, such as indicators of heart disease and mortality risk, while reducing the risk of exposing biometric and demographic information, including age and sex. The system uses advanced neural network architectures to separate clinically relevant signals from identifiable personal characteristics.

Published in Scientific Reports, the study highlights the model’s ability to predict outcomes such as left ventricular ejection fraction (LVEF) while limiting the disclosure of personal information. Researchers report that the system performs competitively compared with existing machine-learning approaches, while improving privacy safeguards.

The researchers also emphasised the importance of reducing bias and improving the representativeness of medical AI systems. Future plans include testing the model across more diverse datasets and releasing it publicly to support safer sharing of ECG data between healthcare institutions.

Why does it matter?

The development might be a critical turning point in medical AI, where improving diagnostic accuracy must be balanced with safeguarding highly sensitive patient information.

As healthcare systems increasingly rely on AI-driven analysis of ECGs and other clinical data, the ability to prevent unintended identification of individuals becomes essential for maintaining trust, enabling secure cross-institutional data sharing, and ensuring compliance with privacy standards.

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

New benchmark tests AI on unpublished mathematics problems

AI systems have demonstrated growing capabilities in advanced mathematics, according to benchmark results published by the non-profit organisation First Proof.

The organisation evaluated four frontier AI systems, including ChatGPT 5.5 Pro, against ten unpublished research-level mathematical problems contributed by leading mathematicians.

The benchmark found that seven of the ten problems received at least one solution judged to be correct by expert reviewers across the participating systems. One notable result involved a stochastic partial differential equations problem, where an AI system produced a correct solution using an approach different from the human-developed proof, drawing praise from expert referees for its originality.

Despite the progress, significant limitations remain.

Several problems remained unsolved, including a metric geometry challenge on which none of the systems made meaningful progress. Reviewers also reported that AI systems handled routine mathematical reasoning effectively but continued to struggle with the most challenging conceptual and creative aspects of proof construction.

Why does it matter?

The benchmark offers one of the most demanding independent tests of AI performance in advanced mathematics, a field often viewed as a proxy for higher-level reasoning and scientific problem-solving. The results suggest that frontier AI systems are increasingly capable of contributing to specialised research tasks and, in some cases, generating approaches that differ from those developed by human experts.

At the same time, the findings highlight the limits of current AI systems. While they can assist with complex reasoning and formal problem-solving, they continue to struggle with the deepest conceptual challenges that often drive mathematical breakthroughs. This suggests that AI may increasingly serve as a research assistant and discovery tool, while human expertise remains essential for guiding and validating scientific advances.

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

INTERPOL report warns of rising cybercrime across Asia-Pacific

INTERPOL has published its 2025/2026 Asia and South Pacific Cyberthreat Assessment Report, covering the period from January 2024 to March 2025. The report documents a rise in cybercrime across the region, attributing the trend to expanding digital infrastructure, the adoption of new technologies and increasingly organised criminal networks.

More than half of the countries surveyed reported that cybercrime accounts for over 30% of all crimes recorded nationally. Phishing and related online scam techniques were identified as the most common and financially damaging forms of cybercrime, with 33 % of surveyed countries recorded over 10,000 such cases.

Neal Jetton, INTERPOL’s Cybercrime Director, said the findings demonstrate how cybercriminals are increasingly exploiting AI, ransomware-as-a-service models and sophisticated social engineering techniques. He noted that operational cooperation, information sharing, and cyber resilience are factors relevant to protecting communities and infrastructure as digital adoption in the region increases.

Growth in internet connectivity, mobile banking, cloud computing, and digital financial services has accompanied this cybercriminal activity, according to the report.

Survey respondents also highlighted challenges for law enforcement, including gaps in specialised forensic tools, cybercrime training and technical capacity. The report also notes differences in cybersecurity capacity across countries.

Some countries have established cybersecurity frameworks and institutional capabilities, while others, including developing countries and small island states, reported resource and capacity constraints.

The report identifies jurisdictions with fragmented enforcement structures, limited technical capabilities, and weaker legislation as more exposed to exploitation by cybercriminal actors.

The report was prepared through the Asia and South Pacific Joint Operations against Cybercrime (ASPJOC) project, funded by the United Kingdom’s Foreign, Commonwealth & Development Office (FCDO). It draws on information submitted by 18 INTERPOL member countries in the Asia and South Pacific region, along with contributions from private sector partners, operational case studies, and analysis of emerging cyber threat trends.

It is one of several regional cyber threat assessments produced by INTERPOL, alongside similar reports covering regions such as Africa. The full report is available from INTERPOL.

Why does this matter?

The report highlights how cybercrime is becoming a major security, economic and governance challenge across Asia and the South Pacific. As countries expand digital infrastructure, online banking, cloud services and digital government initiatives, cybercriminals are finding new opportunities to exploit vulnerabilities and target individuals, businesses and critical sectors.

The findings also illustrate the growing role of AI in cyberspace. While organisations increasingly use AI to strengthen cybersecurity, threat actors are adopting the same technologies to enhance phishing campaigns, generate deepfakes and automate attacks. This accelerating technological competition underscores the importance of international cooperation, cyber capacity-building and information sharing to strengthen resilience across the region.

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

ChatGPT set to join Pentagon’s GenAI.mil platform

Mohammed Husain, OpenAI’s Strategic Delivery Lead for Cyber, said at the Defense One Tech Summit in Virginia that the company expects to launch ChatGPT on GenAI.mil, the US Department of Defense’s enterprise-wide generative AI platform, in early July. The deployment would extend ChatGPT access to more than 3 million defence, civilian, and military personnel.

According to Husain, the version of ChatGPT deployed on GenAI.mil will be certified to handle Controlled Unclassified Information (CUI) and operate at Impact Level 5 (IL5), a Defense Department cloud security classification for systems processing sensitive unclassified information. Husain said OpenAI continues to coordinate with the Pentagon’s Chief Digital and Artificial Intelligence Office (CDAO) on the rollout.

The Department of Defense launched GenAI.mil in December 2025, initially centred on Gemini for Government, before announcing plans to integrate models from OpenAI and xAI. Outside GenAI.mil, federal agencies have had access to ChatGPT since at least January 2025 through ChatGPT Gov.

In August 2025, OpenAI and the General Services Administration reached a OneGov agreement that reduced the price of ChatGPT access for federal agencies. Most recently, OpenAI’s GPT-5.4 model became available to federal government users on Amazon Bedrock and AWS GovCloud earlier this month.

Husain said that as the Department of Defense adopts more capable models, token consumption, the units used by AI systems to process and generate information, is likely to increase, particularly for higher-value tasks.

He pointed to Amazon’s early June announcement that OpenAI’s GPT-5.5, GPT-5.4, and Codex models are now available on Amazon Bedrock as an example of broader access to more capable, token-intensive models.

Husain said token efficiency, measured by the cost of completing tasks rather than raw processing speed, is expected to become an increasingly important consideration in government AI deployments as model capabilities advance.

Why does this matter?

The planned rollout highlights how frontier AI models are moving from experimental deployments into core government and defence infrastructure. Rather than relying on a single provider, the Pentagon is building an ecosystem that includes models from OpenAI, Google and xAI, reflecting a broader strategy of integrating commercial AI capabilities into operational environments.

The development also illustrates the growing institutionalisation of relationships between leading AI companies and national security organisations. As advanced AI systems become embedded in government workflows, questions around security, procurement, oversight, interoperability, and strategic dependence on private-sector AI providers are likely to become increasingly important.

The deployment of ChatGPT on GenAI.mil, therefore, represents not only a technology upgrade but also a step in the evolving governance of AI within national security institutions.

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

European consortium launches SHIELD-6G project to develop cybersecurity capabilities for future 6G networks

A consortium of 19 organisations from across Europe has launched SHIELD-6G (Scalable, Hybrid, and Intelligent End-to-End Defense for 6G Networks), a research and innovation project aimed at developing cybersecurity technologies for future 6G communications networks.

The project is coordinated by University College Dublin and brings together universities, research institutes, telecommunications operators, technology companies, and small and medium-sized enterprises from 10 European countries, including Ireland, Spain, Finland, France, the Netherlands, Italy, Greece, Latvia, Estonia, and Türkiye.

According to the consortium, SHIELD-6G will focus on developing a cyber threat intelligence platform designed for future 6G environments. The platform is intended to support the detection, analysis, and response to cyber threats, including previously unknown vulnerabilities and attacks.

The project will explore several technology areas, including AI-based threat detection and response, federated learning for privacy-preserving data processing, digital twin technologies for security testing, and explainable AI approaches intended to improve transparency in cybersecurity operations.

Researchers will evaluate the technologies through use cases in healthcare, smart manufacturing, and maritime communications. These sectors are expected to rely increasingly on advanced connectivity and automated digital systems, creating new cybersecurity requirements.

The initiative is funded through the European Union’s Horizon Europe programme under the Smart Networks and Services Joint Undertaking (SNS JU), which supports research and innovation activities related to future communication networks and services.

According to the project description, SHIELD-6G is expected to contribute to the development of automated network security capabilities, real-time threat detection and mitigation mechanisms, and approaches to compliance and auditing. The consortium also plans to contribute to ongoing discussions on 6G standardisation.

Commenting on the launch, Madhusanka Liyanage of University College Dublin said future communication networks will require security and resilience measures capable of supporting increasingly critical digital services. He said the project aims to develop cybersecurity capabilities that can help protect those services while supporting the broader development of future connectivity infrastructure.

SHIELD-6G is one of several projects funded under the SNS JU programme that aim to advance research on 6G technologies and related cybersecurity challenges as Europe prepares for the next generation of digital communications networks.

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