Digital standards

Altman proposes US-led international forum for AI safety standards

OpenAI CEO Sam Altman has called for the creation of a US-led international forum to establish global safety standards for AI, arguing that no single country or company should dominate the governance of increasingly capable AI systems.

Writing in an opinion article published in the Financial Times, Altman proposed an international body bringing together governments, independent technical experts, and other stakeholders to develop accepted AI safety standards, provide impartial assessments of AI capabilities and risks, and make advanced AI technologies available to countries and organisations that participate in and comply with agreed rules.

According to Altman, such a forum could also serve as a governance mechanism for frontier AI developers, helping to reduce commercial pressures that may encourage companies to prioritise rapid deployment over safety. He argued that international cooperation has previously enabled countries to manage other strategically important technologies despite geopolitical competition.

To illustrate his proposal, Altman pointed to existing international governance mechanisms such as the International Atomic Energy Agency (IAEA), which oversees the peaceful use of nuclear technology, as well as global aviation safety frameworks and international financial standards. In his view, these models demonstrate that countries can establish common rules for technologies with significant cross-border implications while maintaining national interests.

Altman also argued that the benefits of AI should be shared more broadly, writing that ‘everyone on Earth should benefit from this technology and determine for themselves how best to use it.’ His proposal follows discussions at the recent Group of Seven (G7) summit in France, where executives from OpenAI, Anthropic, and Google DeepMind met with political leaders to discuss international approaches to governing advanced AI models.

A key challenge for any international oversight mechanism, however, remains enforcement. Unlike nuclear facilities or aircraft, frontier AI models are developed within highly secured data centres, making independent verification considerably more difficult. The limited visibility into model training, testing, and deployment has led many experts to question how compliance with international AI standards could be monitored in practice.

Altman’s proposal is not the first call for stronger international oversight of advanced AI. OpenAI and Anthropic have previously supported the idea of international governance mechanisms for frontier AI systems. Anthropic CEO Dario Amodei has argued for a more prescriptive regulatory approach, drawing comparisons with the US Federal Aviation Administration and advocating stronger regulatory oversight for highly capable AI models.

The proposal also comes as governments continue to expand their involvement in AI governance. Alongside national regulatory initiatives, international discussions have accelerated through forums such as the G7, the Global Partnership on AI (GPAI), and the UN.

Earlier this week, the UN’s Independent International Scientific Panel on Artificial Intelligence published its first preliminary assessment of AI opportunities, risks, and governance challenges ahead of the inaugural Global Dialogue on AI Governance in Geneva, reflecting growing international efforts to establish evidence-based approaches to AI governance.

Whether Altman’s proposal develops into a formal international initiative will ultimately depend on governments rather than AI companies. Commenting on broader discussions around AI governance, analysts at the Brookings Institution argued that cooperation between governments and leading AI developers could help establish common standards, but stressed that any future international framework would need effective implementation and enforcement mechanisms rather than relying solely on voluntary commitments.

As governments, international organisations, and AI developers continue debating how to govern increasingly capable AI systems, Altman’s proposal adds to a growing conversation about whether existing institutions are sufficient or whether new international mechanisms will be needed to manage the opportunities and risks associated with frontier AI.

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

Switzerland sets framework for responsible AI use in development co-operation

An OECD case study has highlighted Switzerland’s efforts to govern the responsible use of AI in international cooperate and humanitarian assistance, focusing on a framework adopted by the Swiss Agency for Development and Cooperation in 2025.

The case study says SDC’s AI initiatives had previously been scattered, especially at the country level, while many staff had limited experience with AI. The agency also lacked unified guidance for using AI tools, funding AI-related projects and engaging in policy dialogue.

Approved in 2025, SDC’s Working Aid on AI is grounded in Switzerland’s International Cooperation Strategy 2025–2028. It provides practical guidance for responsible AI adoption across the agency’s portfolio and institutional roles.

The framework draws on earlier risk and opportunity mapping, the Council of Europe’s Framework Convention on AI and the OECD AI principles.

Its guiding principles include doing no harm, human oversight, participation of affected communities, localisation, dataset debiasing, ethical data sourcing, decent work in AI supply chains, reduced climate impact, transparency and internal oversight.

The Working Aid also defines four roles for SDC: funding operational AI projects, influencing global AI policy and partnerships, providing sectoral advice to SDC units and Swiss representations, and embedding AI into knowledge management.

SDC has created an AI Task Force, now becoming an AI Network, to coordinate work on operations, staff skills, data and IT infrastructure, governance and partnerships.

The framework is already being applied to areas including climate forecasting, child health diagnostics, media development, disinformation and internal project-cycle management.

Why does it matter?

Switzerland’s approach shows how development agencies are beginning to institutionalise AI governance rather than treating AI as a series of isolated experiments. A framework for responsible use can help agencies manage risks around bias, dependency, data sourcing, climate impact and human oversight while still using AI for development and humanitarian goals. The case also highlights the importance of internal capacity, staff guidance and whole-of-government coordination as AI becomes part of international cooperate.

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

Ericsson report says global 5G subscriptions pass 3 billion

Global 5G subscriptions passed 3 billion in the first quarter of 2026, according to Ericsson’s latest Mobility Report.

The report says 162 million 5G subscriptions were added during the quarter, bringing the global total to 3.1 billion. Ericsson expects 5G subscriptions to more than double to 6.4 billion by the end of 2031.

5G will also carry around half of global mobile data traffic by the end of 2025. Ericsson projects that 5G networks will account for 85% of mobile data traffic by 2031.

The report highlights the continued deployment of 5G Standalone networks and the growth of commercial network slicing services, which allow operators to offer differentiated connectivity for specific use cases.

Ericsson also points to changing traffic patterns. For many service providers, uplink traffic is already growing faster than downlink traffic, driven by collaboration tools, cloud storage and emerging services that require more data to be sent from devices to networks.

The company says AI-powered devices, augmented reality applications and connected technologies are likely to increase demand for real-time data processing and uplink capacity.

Ericsson said existing 5G networks can support early AI and extended reality services, while 6G is expected to enable larger-scale AI-native applications, with the first commercial services expected around 2030.

Why does it matter?

The report shows that 5G is becoming a core layer of digital infrastructure for AI-enabled services, cloud applications and connected devices. As AI moves from centralised data centres into devices, vehicles, workplaces and industrial systems, mobile networks will need to support higher uplink capacity, lower latency and more differentiated connectivity. Growth in 5G Standalone and network slicing also matters because these technologies give operators more tools to support specialised services, from enterprise automation to future AI and XR applications.

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

OECD explores AI-powered regulatory inspections

The Organisation for Economic Co-operation and Development (OECD) has published a working paper examining how data-driven regulation and digital technologies, including AI and data analytics, can help authorities carry out more targeted, risk-based and effective inspections.

The paper identifies licensing, permitting and inspections as the three pillars of regulatory delivery, arguing that these mechanisms are most effective when supported by risk-based approaches that minimise unnecessary administrative burdens while improving regulatory outcomes. The core argument is that by adopting risk-based approaches supported by technology, regulators can concentrate their efforts where they are most needed rather than applying uniform enforcement across all actors.

The OECD highlights practical uses for AI and data analytics, including identifying high-risk areas, prioritising inspections, streamlining enforcement and allocating resources more efficiently. The aim is to improve compliance while reducing unnecessary interventions for lower-risk businesses and activities.

The paper also argues that technologies can strengthen public trust in regulation by making inspections more transparent, consistent and evidence-based, improving both the effectiveness and legitimacy of regulatory enforcement.

The project forms part of broader EU efforts to modernise regulatory delivery. Drawing on Italy’s pilot experience, the OECD aims to identify lessons that can be applied across member states and other jurisdictions pursuing evidence-based regulatory reform.

Why does it matter?

The paper illustrates how AI and data analytics could help regulators move away from one-size-fits-all enforcement towards more targeted, risk-based oversight. By focusing inspections where they are most needed, authorities could improve compliance while reducing unnecessary administrative burdens, particularly for smaller businesses.

The report also reflects a wider shift towards evidence-based regulation. As governments seek to modernise public administration without weakening regulatory standards, technologies such as AI are increasingly being viewed as tools for improving both regulatory efficiency and public trust through more transparent and proportionate enforcement.

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

Anthropic redeploys Claude Fable 5

Anthropic will restore global access to Claude Fable 5 after the US government lifts export controls on the model.

The company said the controls were applied on 12 June to Claude Fable 5 and Claude Mythos 5, requiring access restrictions for foreign nationals inside and outside the United States. Anthropic suspended access to both models for all users because it said it had no reliable way to verify nationality in real time.

Anthropic said the controls were lifted on 30 June. Fable 5 will become available globally from 1 July on the Claude Platform, Claude.ai, Claude Code and Claude Cowork, with access on AWS, Google Cloud and Microsoft Foundry to be restored as quickly as possible.

Access to Mythos 5 has been restored only for a set of US organisations following government approval. Anthropic said Fable 5 and Mythos 5 share the same underlying model, but Fable 5 has stronger safeguards for general use, while Mythos 5 has fewer safeguards and is limited to trusted partners working on defensive cybersecurity.

The export control directive followed a report by Amazon researchers describing a method for bypassing Fable 5 safeguards. Anthropic said the reported behaviour involved identifying software vulnerabilities and, in one case, producing code showing how a vulnerability could be exploited.

The company said its review found that the technique did not expose unique Mythos-level cyber capabilities. It has trained an improved safety classifier to block the behaviour described in the report, and said blocked requests will be redirected to Claude Opus 4.8.

Anthropic also called for a shared industry framework to assess the severity of AI jailbreaks. It said it is working with Amazon, Microsoft, Google and other Glasswing partners on criteria including capability gain, breadth of capability gain, ease of weaponisation and discoverability.

The company said it is expanding cooperation with the US government on frontier AI security, including pre-release evaluation, faster information sharing and joint research on safeguards.

Why does it matter?

The case shows how frontier AI releases are becoming part of national security and export-control policy, especially when models have advanced cybersecurity capabilities. Anthropic’s response also highlights a broader governance gap: governments and companies still lack a shared standard for judging when a jailbreak is minor, serious or urgent enough to justify intervention. The outcome could influence how advanced AI models are tested, released and restricted across borders.

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

EPO reports record patent demand as AI and digital services boost innovation

The European Patent Office (EPO) has published its Annual Review 2025, revealing that European patent applications exceeded 200,000 for the first time in the organisation’s history.

The milestone reflects growing confidence in the European patent system, supported by continued investment in digital transformation, AI and more efficient patent examination processes under the Strategic Plan 2028.

The Office processed a record 418,868 patent dossiers during 2025, increasing productivity by 4% while maintaining high quality standards and improving the speed of patent searches, grants and opposition proceedings.

User satisfaction also remained high following the EPO’s largest-ever satisfaction survey, involving more than 8,000 participants. Innovation activity continued to grow across strategic sectors including digital technologies, healthcare, advanced materials and battery technologies.

AI played an increasingly important role throughout the patent granting process. The EPO expanded AI-powered tools for patent examiners, including a large language model-based enhancement to its PreSearch system, designed to improve prior art discovery while ensuring examiners retain full control over decision-making.

Additional AI-supported capabilities now assist with document analysis, advanced searches, file allocation and oral proceedings. At the same time, MyEPO continued evolving as the organisation’s central digital platform, while Online Filing 2.0 became the standard filing tool ahead of broader DOCX filing deployment.

The report also highlights the growing success of the Unitary Patent system, with SMEs, universities and public research organisations accounting for nearly half of all Unitary Patents granted to European innovators.

Alongside new innovation intelligence tools such as the Patent Standards Explorer, Digital Library and expanded Deep Tech Finder, the EPO says it is strengthening Europe’s innovation ecosystem through greater transparency, digital services and data-driven patent intelligence.

Why does it matter?

The Annual Review demonstrates how AI is becoming embedded within one of Europe’s most important innovation institutions. Rather than replacing patent examiners, AI is being deployed to improve search quality, accelerate administrative processes and strengthen decision-making while maintaining human oversight.

It also illustrates Europe’s broader strategy of combining AI adoption with digital public services, intellectual property protection and innovation policy.

Record patent demand, expanding use of the Unitary Patent and new digital tools suggest the EPO is positioning itself as a key pillar of Europe’s competitiveness in emerging technologies, particularly as global competition intensifies in AI, semiconductors, advanced manufacturing and deep tech.

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

UK CMA consults on Apple and Google app store payment rules

The UK Competition and Markets Authority has opened consultations on new requirements for Apple and Google under the country’s digital markets competition regime.

Proposed steering requirements would allow app developers to direct UK users to payment options outside Apple’s App Store and Google’s Play Store. The CMA said Apple currently bans steering in the UK, while Google restricts it.

According to the regulator, allowing developers to communicate with customers about off-platform options could increase competition, reduce payment costs and support innovation across mobile services.

Consultation proposals also cover principles to ensure that any steering fees charged by Apple and Google are fair and reasonable. The CMA said such fees should be based on evidence and should be lower than current app store charges.

Alongside the steering proposals, officials are seeking views on a potential requirement for Apple to provide developers with access to near-field communication functionality on iOS.

Broader NFC access could allow UK fintechs and developers to support contactless payments from within their own apps. It could also support future payment methods, including account-to-account payments, digital currencies and stablecoins, as well as non-financial uses such as digital ID and car keys.

Responses to the steering conduct requirement are due by 28 July 2026, while views on the potential NFC requirement are due by 21 July 2026. The CMA will decide later this year whether to impose new obligations.

Why does it matter?

The consultations show the UK’s digital markets regime moving into targeted conduct rules for major mobile platforms. If adopted, the measures could weaken Apple and Google’s control over in-app payments and give developers more freedom to offer alternative purchasing channels. The NFC proposal also widens the debate beyond app store commissions, addressing Apple’s control over device functionality that can shape competition in mobile payments, digital identity and other services.

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

Spain’s AI sandbox offers early test for biometric AI compliance

Spain’s AI regulatory sandbox is becoming an early test of how high-risk AI systems may prepare for compliance with the EU AI Act, with facial recognition among the technologies examined.

Spanish company Herta said it has completed the sandbox process for its facial-recognition video-surveillance system, BioSurveillance. The company presented the pilot as a step towards AI Act-ready deployments in public settings.

Herta describes BioSurveillance as a real-time video-surveillance system capable of detecting multiple faces, enrolling individuals during operation, identifying previously registered people and managing alerts. Its BioSurveillance NEXT product is designed for simultaneous identification in crowded and changing environments.

Spain’s AI agency, AESIA, says practical guides developed through the national AI regulatory sandbox are intended to help companies that develop or deploy high-risk AI systems prepare for their obligations under the EU AI Act. The guides provide recommendations while harmonised EU standards are still being developed.

However, sandbox participation should not be treated as approval for public facial recognition deployments. Remote biometric identification in publicly accessible spaces remains one of the most sensitive areas under the EU AI Act. It is subject to strict limits, depending on the use case, operator and context.

The case highlights how companies developing biometric AI systems are seeking early compliance pathways, while regulators face pressure to balance innovation, public safety, privacy and fundamental rights.

Why does it matter?

Facial recognition is one of the most contested areas of AI regulation because it combines public-space surveillance, biometric data processing and risks to privacy and fundamental rights. Spain’s sandbox offers an early view of how high-risk AI providers may prepare documentation, testing and compliance processes under the EU AI Act. The case also shows why compliance language must be used carefully: participation in a sandbox may support readiness, but it does not remove the legal restrictions surrounding biometric identification in public spaces.

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

MIT develops AI system to improve robot understanding

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory have developed a system that helps robots interpret vague human instructions while using significantly less training data.

The approach, called Masked Inverse Reinforcement Learning (Masked IRL), uses two large language models to clarify tasks and identify the details that matter for safe robot movement.

One model expands ambiguous instructions based on user demonstrations. A second model filters out irrelevant information and highlights factors the robot should include in its motion plan.

The system can help robots understand unstated preferences, such as avoiding a laptop while delivering a coffee mug or keeping a safe distance from a person during a task.

MIT said Masked IRL correctly identified users’ unstated preferences up to 15% more often than comparable methods. Researchers also found that it required nearly five times less demonstration data to learn new tasks.

The approach was tested in simulated environments and on a real robotic arm. The robot completed tasks it had not seen during training, including moving a cup towards a person while avoiding a computer and handing over an object while staying away from nearby obstacles.

Researchers plan to make the system more dynamic by adding cameras, enabling robots to identify relevant objects and ignore distractions in their surroundings visually.

Why does it matter?

Masked IRL could make robots easier to deploy in homes, offices, factories and care environments by reducing the amount of human training needed. The system also addresses a core safety challenge in robotics: people often give vague instructions and leave important preferences unstated. Better interpretation of human intent could help robots work more safely around people, objects and changing environments.

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

Australian audit highlights governance gaps in public-sector AI

The Australian National Audit Office has found that IP Australia’s use of AI in the patent rights process is largely effective, while calling for stronger cybersecurity governance, monitoring and strategic oversight.

Auditor-General Report No. 43 of 2025–26 examined whether IP Australia has effective arrangements to support AI adoption in the patent rights process. IP Australia administers intellectual property rights, including patents, trade marks, design rights and plant breeders’ rights.

The agency deployed its first AI tool for patent examination in 2018 and now uses four AI tools in the process. The tools are designed to provide examiners with information to support better decisions, rather than to decide patent applications themselves.

The ANAO said IP Australia has been an early adopter of AI and has progressively improved its governance arrangements. The agency has introduced an AI governance policy, risk-scaled assessment mechanisms and clearer enterprise accountability roles.

However, the audit found that strategic oversight of AI implementation and related benefits is not yet fully established. It said IP Australia’s AI inventory, committee roles and use-case ownership remain works in progress.

Monitoring and reporting were assessed as only partly effective. The ANAO said benefits have been inconsistently defined and measured, making it harder to demonstrate the ongoing effectiveness of AI tools and manage emerging risks.

The ANAO made two recommendations, urging IP Australia to review cybersecurity governance controls for AI and establish clearer risk-based monitoring and reporting arrangements. IP Australia agreed to both recommendations.

The audit said public-sector agencies should regularly reassess AI governance frameworks as they move from experimentation to wider use.

Why does it matter?

The audit shows how AI is moving from experimentation into routine public-sector decision support. IP Australia’s experience points to the benefits of AI in improving efficiency and quality, but also shows that governance must evolve as tools become embedded in official processes. Cybersecurity, accountability, monitoring and measurable benefits are becoming central to responsible AI use in government.

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

Diplo chatbot can make mistakes. Consider checking important information.