OpenAI has released a set of prompt-based safety policies to help developers build safer AI experiences for teenagers. The tools work with the open-weight model gpt-oss-safeguard, turning safety requirements into practical classifiers for real-world use.
The policies address teen risks, including graphic violence, sexual content, harmful body image behaviour, dangerous challenges, roleplay, and age-restricted goods and services. Developers can use them for both real-time filtering and offline content analysis.
The framework was developed with input from organisations such as Common Sense Media and everyone.ai to improve clarity and consistency in teen safety rules. The initiative also responds to long-standing challenges in translating high-level safety goals into precise operational systems.
Open-source availability through the ROOST Model Community allows developers to adapt and expand the policies for different use cases and languages. The framework is a foundational step, not a complete solution, encouraging layered safeguards and ongoing refinement.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
A transition to post-quantum cryptography by 2029 is being led by Google, aiming to secure digital systems against future quantum computing threats instead of relying on existing encryption standards.
Quantum computers are expected to challenge widely used encryption and digital signature systems, prompting the need for early transition strategies.
Google has updated its threat model to prioritise authentication services, recognising that digital signatures pose a critical vulnerability if not addressed before the arrival of quantum machines capable of cryptanalysis.
The company is encouraging broader industry action to accelerate migration efforts and reduce long-term security risks.
As part of its strategy, Google is integrating post-quantum cryptography into its products and services.
Android 17 will include quantum-resistant digital signature protection aligned with standards developed by the US’s National Institute of Standards and Technology. At the same time, support has already been introduced in Google Chrome and cloud platforms.
These measures aim to bring advanced security technologies directly to users instead of limiting them to experimental environments.
By setting a clear timeline, Google aims to instil urgency and direction across the wider technology sector.
The transition to post-quantum cryptography is expected to become a critical step in maintaining online security, ensuring that digital infrastructure remains resilient as quantum computing capabilities continue to evolve.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
OpenAI has introduced a public Safety Bug Bounty programme to identify misuse and safety risks across its AI systems. The initiative expands the company’s existing vulnerability reporting framework by focusing on harms that fall outside traditional security definitions.
The programme covers AI threats such as agentic risks, prompt injection, data exfiltration, and bypassing platform integrity controls. Researchers are encouraged to submit reproducible cases where AI systems perform harmful actions or expose sensitive information.
Unlike standard security reports, the initiative accepts safety issues that pose real-world risk, even if they are not classified as technical vulnerabilities. Dedicated safety and security teams will assess submissions and may be reassigned depending on relevance.
The scheme is open to external researchers and ethical hackers to strengthen AI safety through broader collaboration. OpenAI says the approach is intended to improve resilience against evolving misuse as AI systems become more advanced.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Executive Vice-President Henna Virkkunen will host a high-level dialogue in Brussels to assess the implementation of the European Chips Act Regulation and gather industry feedback ahead of its planned revision.
Stakeholders from across the semiconductor ecosystem are expected to exchange views and present recommendations to shape future policy direction.
An initiative that forms part of the broader strategy led by the European Commission to reinforce technological sovereignty and competitiveness, rather than relying heavily on external suppliers.
The Chips Act seeks to strengthen Europe’s semiconductor ecosystem, improve supply chain resilience, and reduce strategic dependencies in critical technologies.
The dialogue follows a public consultation and call for evidence conducted in autumn 2025, with findings set to inform the upcoming legislative revision.
Industry representatives will provide direct input through a report outlining challenges, opportunities, and proposed policy adjustments, contributing to a more targeted and effective framework for semiconductor development.
Looking ahead, the revision of the Chips Act will be integrated into a wider Technological Sovereignty package designed to boost the capacity of Europe’s digital industries.
By combining stakeholder engagement with policy reform, the European Commission aims to ensure that semiconductor innovation and production can expand across the EU rather than remain constrained by reliance on external suppliers.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
The OECD (Organisation for Economic Co-operation and Development) highlights how businesses are preparing for quantum computing, recognising it as a transformative technology instead of relying solely on conventional computing methods.
Quantum readiness is framed as a long-term capability-building effort in which firms gradually develop skills, infrastructure, and partnerships to explore commercial applications while navigating uncertainty.
Drawing on research, surveys, and interviews with public and private organisations across 10 countries, the OECD identifies both the practical steps companies take to build readiness and the barriers that slow adoption.
Early efforts focus on low-cost awareness and exploration, including attending workshops, training sessions, and industry events, allowing firms to familiarise themselves with emerging opportunities instead of waiting for fully mature systems.
Despite growing interest, companies face significant challenges. Technological immaturity complicates pilots and feasibility studies, while many firms lack a clear understanding of potential business applications.
Access to quantum resources, funding for research and development, and staff training are expensive, particularly for small- and medium-sized enterprises. Furthermore, there is a shortage of talent with both quantum computing expertise and domain-specific knowledge.
As a result, readiness tends to be concentrated among large, R&D-intensive firms, while smaller companies often recognise quantum computing’s potential but delay action.
Such an uneven adoption risks creating a divide in the digital economy, with early adopters moving ahead and other firms falling behind instead of engaging proactively.
To address these challenges, the OECD notes that public and private support mechanisms are critical. Networking and collaboration platforms connect firms with researchers, technology providers, and industry peers, fostering knowledge exchange and collective experimentation.
Business advisory and technology extension services help companies assess capabilities, test solutions, and access specialised facilities.
Grants for research and development lower the costs of experimentation and encourage collaboration, while stakeholder consultations ensure that support measures remain aligned with business needs.
Many companies are also establishing internal quantum labs and innovation hubs to trial applications and build expertise in a controlled environment, combining research with practical exploration instead of relying solely on external guidance.
Looking ahead, the OECD recommends expanding education and skills pipelines, strengthening industry-academic partnerships, and designing policies that support broader participation in quantum adoption.
Hybrid approaches that integrate quantum computing with AI and high-performance computing may offer practical commercial entry points for early applications.
Policymakers are encouraged to balance near-term exploratory pilots with forward-looking support for software development, interoperability, and workforce growth, enabling firms to move from experimentation to deployment effectively.
The European Data Protection Board and the European Data Protection Supervisor have backed proposals to strengthen the EU cybersecurity law while safeguarding personal data. Their joint opinion addresses reforms to the Cybersecurity Act and updates to the NIS2 Directive.
Regulators support plans to reinforce the mandate of the European Union Agency for Cybersecurity and expand cybersecurity certification across digital supply chains. Clearer coordination between ENISA and privacy authorities is seen as essential for consistent oversight.
Advice also calls for limits on the processing of personal data and for prior consultation on technical rules affecting privacy. Certification schemes should align with the GDPR and help organisations demonstrate compliance.
Additional recommendations include broader cybersecurity skills training and a single EU entry point for personal data breach notifications. Proposed changes would also classify digital identity wallet providers as essential entities under the EU security rules.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
A major expansion of its activities has been outlined by OpenAI Foundation, signalling a broader effort to ensure AI delivers tangible benefits while addressing emerging risks.
The organisation plans to invest at least $1 billion over the next year, forming part of a wider $25 billion commitment focused on disease research and AI resilience.
OpenAI Foundation frames such potential as central to its mission, while recognising that more capable systems introduce complex societal and safety challenges that require coordinated responses.
Initial programmes prioritise life sciences, including research into Alzheimer’s disease, expanded access to public health data, and accelerated progress on high-mortality conditions.
Parallel efforts examine the economic impact of automation, with engagement across policymakers, labour groups and businesses aimed at developing practical responses to labour market disruption.
A dedicated resilience strategy addresses risks linked to advanced AI systems, including safety standards, biosecurity concerns and the protection of children and young users.
Alongside community-focused funding, the OpenAI Foundation’s initiative reflects a dual objective: enabling innovation rather than leaving societies exposed to technological disruption.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Over the past few years, we have witnessed a rapid shift in the way data is stored and processed across businesses, organisations, and digital systems.
What we are increasingly seeing is that AI itself is changing form as computation shifts away from centralised cloud environments to the network edge. Such a shift has come to be known as edge AI.
Edge AI refers to the deployment of machine learning models directly on local devices such as smartphones, sensors, industrial machines, and autonomous systems.
Instead of transmitting data to remote servers for processing, analysis is performed on the device itself, enabling faster responses and greater control over sensitive information.
Such a transition marks a significant departure from earlier models of AI deployment, where cloud infrastructure dominated both processing and storage.
From centralised AI to edge intelligence
Traditional AI systems used to rely heavily on centralised architectures. Data collected from users or devices would be transmitted to large-scale data centres, where powerful servers would perform computations and generate outputs.
Such a model offered efficiency, scalability, and easier security management, as protection efforts could be concentrated within controlled environments.
Centralisation allowed organisations to enforce uniform security policies, deploy updates rapidly, and monitor threats from a single vantage point. However, reliance on cloud infrastructure also introduced latency, bandwidth constraints, and increased exposure of sensitive data during transmission.
Edge AI introduces a fundamentally different paradigm. Moving computation closer to the data source reduces the reliance on continuous connectivity and enables real-time decision-making.
Such decentralisation represents not merely a technical shift but a reconfiguration of the way digital systems operate and interact with their environments.
Advantages of edge AI
Reduced latency and real-time processing
Latency is significantly reduced when computation occurs locally. Edge systems are particularly valuable in time-sensitive applications such as autonomous vehicles, healthcare monitoring, and industrial automation, where delays can have critical consequences.
Enhanced privacy and data control
Privacy improves when sensitive data remains on-device instead of being transmitted across networks. Such an approach aligns with growing concerns around data protection, regulatory compliance, and user trust.
Operational resilience
Edge systems can continue functioning even when network connectivity is limited or unavailable. In remote environments or critical infrastructure, independence from central servers ensures service continuity.
Bandwidth efficiency and cost reduction
Bandwidth consumption is decreased because only processed insights are transmitted, not raw data. Such efficiency can translate into reduced operational costs and improved system performance.
Personalisation and context awareness
Devices can adapt to user behaviour in real time, learning from local data without exposing sensitive information externally. In healthcare, personalised diagnostics can be performed directly on wearable devices, while in manufacturing, predictive maintenance can occur on-site.
The dark side of edge AI
However, the shift towards edge computing introduces profound cybersecurity challenges. The most significant of these is the expansion of the attack surface.
Instead of a limited number of well-protected data centres, organisations must secure vast networks of distributed devices. Each endpoint represents a potential entry point for malicious actors.
The scale and diversity of edge deployments complicate efforts to maintain consistent security standards. Security is no longer centralised but dispersed, increasing the likelihood of vulnerabilities and misconfigurations.
Let’s take a closer look at some other challenges of edge AI.
Physical vulnerabilities and device exposure
Edge devices often operate in uncontrolled environments, making physical access a major risk. Attackers may tamper with hardware, extract sensitive information, or reverse engineer AI models.
Model extraction attacks allow adversaries to replicate proprietary algorithms, undermining intellectual property and enabling further exploitation. Such risks are significantly more pronounced compared to cloud systems, where physical access is tightly controlled.
Software constraints and patch management challenges
Many edge devices rely on embedded systems with limited computational resources. Such constraints make it difficult to implement robust security measures, including advanced encryption and intrusion detection.
Patch management becomes increasingly complex in decentralised environments. Ensuring that millions of devices receive timely updates is a significant challenge, particularly when connectivity is inconsistent or when devices operate in remote locations.
Breakdown of traditional security models
The decentralised nature of edge AI undermines conventional perimeter-based security frameworks. Without a clearly defined boundary, traditional approaches to network defence lose effectiveness.
Each device must be treated as an independent security domain, requiring authentication, authorisation, and continuous monitoring. Identity management becomes more complex as the number of devices grows, increasing the risk of misconfiguration and unauthorised access.
Data integrity and adversarial threats
As we mentioned before, edge devices rely heavily on local data inputs to make decisions. As a result, manipulated inputs can lead to compromised outcomes. Adversarial attacks, in which inputs are deliberately altered to deceive machine learning models, represent a significant threat.
In safety-critical systems, such manipulation can lead to severe consequences. Altered sensor data in industrial environments may disrupt operations, while compromised vision systems in autonomous vehicles may produce dangerous behaviour.
Supply chain risks in edge AI
Edge AI systems depend on a combination of hardware, software, and pre-trained models sourced from multiple vendors. Each component introduces potential vulnerabilities.
Attackers may compromise supply chains by inserting backdoors during manufacturing, distributing malicious updates, or exploiting third-party software dependencies. The global nature of technology supply chains complicates efforts to ensure trust and accountability.
Energy constraints and security trade-offs
Edge devices are often designed with efficiency in mind, prioritising performance and power consumption. Security mechanisms such as encryption and continuous monitoring require computational resources that may be limited.
As a result, security features may be simplified or omitted, increasing exposure to cyber threats. Balancing efficiency with robust protection remains a persistent challenge.
Cyber-physical risks and real-world impact
The integration of edge AI into cyber-physical systems elevates the consequences of security breaches. Digital manipulation can directly influence physical outcomes, affecting safety and infrastructure.
Compromised healthcare devices may produce incorrect diagnoses, while disrupted transportation systems may lead to accidents. In energy networks, attacks could impact entire regions, highlighting the broader societal implications of edge AI vulnerabilities.
Regulatory and governance challenges
Existing regulatory frameworks have been largely designed for centralised systems and do not fully address the complexities of decentralised architectures. Questions regarding liability, accountability, and enforcement remain unresolved.
Organisations may struggle to implement effective security practices without clear standards. Policymakers face the challenge of developing regulations that reflect the distributed nature of edge AI systems.
Towards a secure edge AI ecosystem
Addressing all these challenges requires a multi-layered and adaptive approach that reflects the complexity of edge AI environments.
Hardware-level protections, such as secure enclaves and trusted execution environments, play a critical role in safeguarding sensitive operations from physical tampering and low-level attacks.
Encryption and secure boot processes further strengthen device integrity, ensuring that both data and models remain protected and that unauthorised modifications are prevented from the outset.
At the software level, continuous monitoring and anomaly detection are essential for identifying threats in real time, particularly in distributed systems where central oversight is limited.
Secure update mechanisms must also be prioritised, ensuring that patches and security improvements can be deployed efficiently and reliably across large networks of devices, even in conditions of intermittent connectivity.
Without such mechanisms, vulnerabilities can persist and spread across the ecosystem.
Rather than relying entirely on decentralised or centralised models, organisations are distributing workloads strategically, keeping latency-sensitive and privacy-critical processes on the edge while maintaining centralised oversight, analytics, and security coordination in the cloud.
Such an approach allows organisations to balance performance and control, while enabling more effective threat detection and response through aggregated intelligence.
Security must also be embedded into system design from the outset, rather than treated as an additional layer to be applied after deployment. A proactive approach to risk assessment, combined with secure development practices, can significantly reduce vulnerabilities before systems are operational.
In conclusion, we have seen how the rise of edge AI represents a pivotal shift in both AI and cybersecurity. Decentralisation enables faster, more private, and more resilient systems, yet it also creates a fragmented and dynamic attack surface.
The advantages we have outlined are compelling, but they also introduce additional layers of complexity and risk. Addressing these challenges requires a comprehensive approach that combines technological innovation, regulatory development, and organisational awareness.
Only through such coordinated efforts can the benefits of edge AI be realised while ensuring that security, trust, and safety remain intact in an increasingly decentralised digital landscape.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
Binance has launched the Beta version of Binance Ai Pro, an advanced AI trading assistant built on the OpenClaw ecosystem. Available from 25 March 2026 at 07:00 UTC, the service can be activated via the Binance App on Android or through the Binance web homepage, with iOS support coming soon.
The platform offers one-click activation, automatic cloud setup, and integration with multiple AI models, including ChatGPT, Claude, Qwen, MiniMax, and Kimi. Users receive a dedicated Binance Ai Pro Account, isolated from their main account to minimise operational risks.
Funds can be manually transferred to the AI account for trading, asset monitoring, and strategy execution, covering spot and perpetual contracts, leveraged borrowing, market analysis, token distribution queries, and custom strategies.
Beta users will pay $9.99 per month, with a 7-day free trial. Activation grants 5 million usage credits each month for accessing advanced AI models, with automatic fallback to basic models once credits are exhausted.
Security measures ensure that AI API keys have no withdrawal permissions and operate within strict, authorised scopes.
Binance plans to expand the platform with additional credits, enriched Binance Skills, and user-customisable third-party AI tools. The company warns that AI trading carries risks and urges users to trade responsibly while giving feedback to enhance the platform.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
The European Commission and Australia have announced the adoption of a Security and Defence Partnership alongside the conclusion of negotiations for a free trade agreement.
They have also agreed to launch formal negotiations for Australia’s association with Horizon Europe, the European Union’s research and innovation funding programme.
The Security and Defence Partnership establishes a framework for cooperation on shared strategic priorities. It includes coordination on crisis management, maritime security, cybersecurity, and countering hybrid threats and foreign information manipulation.
A partnership that also includes cooperation on emerging and disruptive technologies, including AI, as well as space security, non-proliferation, and disarmament.
The free trade agreement provides for the removal of over 99% of tariffs on the EU goods exports to Australia and expands access to services, government procurement, and investment opportunities.
It includes provisions on data flows that prohibit data localisation requirements and supports supply chain resilience through improved access to critical raw materials.
The EU exports are expected to increase by up to 33% over the next decade.
The agreement incorporates commitments on trade and sustainable development, including labour rights, environmental standards, and climate obligations aligned with the Paris Agreement.
The negotiated texts will undergo the EU internal procedures before submission to the Council for signature and conclusion, followed by European Parliament consent and ratification by Australia before entry into force.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
