Artificial intelligence Archives | Page 3 of 708

Meta adds suicide prevention safeguards to AI chats for teens

Meta has announced new safety measures for teenagers using Meta AI, including parental alerts when supervised teens show signs of suicide or self-harm during conversations with the chatbot. The company said alerts will be sent only after manual review and will include guidance to help parents support their child.

The company is also developing a system to notify emergency services when AI conversations indicate someone may face an imminent risk of suicide. The approach builds on the company’s existing practice of referring serious suicide risks identified on Facebook and Instagram to emergency responders.

Meta said it worked with more than 75 mental health clinicians to improve how its AI responds when teenagers discuss suicide or self-harm. The updated responses are intended to acknowledge users’ feelings while directing them towards appropriate offline support and crisis services.

The company is also extending its stricter ‘Limited Content’ setting to Meta AI chats, restricting a wider range of sensitive conversations for supervised teens. The parental alerts are now available in the US, UK, Australia and Canada, with global rollout planned by the end of the year.

Why does it matter?

The measures illustrate how AI safety is expanding beyond preventing harmful content to managing situations in which users may face immediate risks to their wellbeing. As conversational AI becomes more widely used by teenagers, developers are increasingly expected to incorporate safeguards, human oversight and access to professional support into their systems.

The announcement also highlights the growing convergence between AI governance and child online safety. Features such as parental notifications, clinically informed responses and emergency escalation suggest that AI assistants are beginning to adopt safety frameworks previously developed for social media platforms, raising new questions about privacy, duty of care and appropriate intervention.

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

Researchers demonstrate prompt injection attacks on Gemini

Kaspersky security researchers have demonstrated new attack techniques that could manipulate Google’s Gemini AI assistant into performing unauthorised actions via prompt injection. The study found that malicious instructions hidden in calendar invites, emails or text messages could bypass safeguards by exploiting how large language models process information.

According to the research, attackers could combine indirect prompt injection, memory poisoning and delayed execution to influence Gemini’s behaviour. Potential outcomes include sending emails, launching applications, controlling compatible smart home devices, displaying false information or embedding malicious instructions into the assistant’s long-term memory, provided the user has granted the necessary permissions.

Researchers also warned that smartphones significantly expand the potential attack surface because Gemini can access notifications containing SMS messages, instant messages and social media alerts. Although Google has addressed the specific vulnerabilities identified in the research, the report argues that prompt injection remains a fundamental challenge for AI systems and that new attack methods are likely to emerge.

The researchers recommend reducing Gemini’s access to notifications, connected apps and system functions where possible, turning off unnecessary AI features and limiting permissions to minimise the impact of future attacks. They also advise users to review AI assistant settings regularly as security protections continue to evolve.

Why does it matter?

Prompt injection represents a major challenge for AI security because it targets how large language models interpret and prioritise information. As AI assistants gain broader access to personal data and connected devices, stronger safeguards will be needed to reduce potential risks.

The research highlights the importance of developing more robust AI security frameworks, including improved permission management and continuous testing, to ensure reliable and responsible adoption of AI technologies.

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

Nearly six in ten UK adults now use generative AI, survey finds

Nearly six in ten UK adults used generative AI during the previous three months, according to the Department for Science, Innovation and Technology’s (DSIT) first Public Engagement Survey.

The survey found that 59% of adults had used generative AI, while 97% reported at least some awareness of AI. Tools capable of producing human-like text or speech were used by 56% of respondents, while 54% had used AI-powered digital assistants.

Adoption was highest among adults aged 16 to 44 before declining steadily with age, reaching just 8% among people aged 85 and over. Use also increased with education levels, from 19% among adults without formal qualifications to 75% among degree holders. Regionally, London recorded the highest adoption rate at 69%, compared with 51% in North East England.

Among online respondents using the technology, 83% reported personal use, 54% used it for work, and 27% for education or study. Saving time was the main motivation, followed by generating ideas and summarising information. Although easier access to information or advice was seen as AI’s biggest benefit, 74% of adults remained concerned that AI-generated information could be inaccurate.

The findings also point to broad public support for regulation. At least two-thirds of respondents backed legislation to improve AI safety, increase transparency, reduce bias and limit job displacement.

The survey was conducted between November 2025 and March 2026 and included 30,698 respondents, providing one of the broadest official snapshots of generative AI use across the UK.

Why does it matter?

The survey suggests that generative AI has moved rapidly into the mainstream, making questions of governance, digital skills and public trust increasingly important. As adoption grows across everyday life and work, ensuring people can use AI safely and effectively may become as important as expanding access to the technology itself.

The findings also highlight persistent inequalities in adoption across age, education and geography. Combined with strong public support for AI regulation, they suggest that future policy will need to address both access to AI and confidence in how it is developed and used.

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

China promotes inclusive AI governance at Shanghai conference

China has pushed back against claims that its open-source AI strategy and cooperation with developing countries are intended to expand its geopolitical influence.

Foreign Ministry spokesperson Lin Jian said on 16 July that AI development affects all humanity and should be guided by openness, inclusion and stronger international governance.

He called for efforts to narrow digital and technological divides, strengthen AI safety and give developing countries greater access to the benefits of the technology.

Lin also criticised technological restrictions and attempts to divide AI development along ideological lines, saying China supports sharing experience and expanding global AI innovation.

The comments came ahead of the World AI Conference and High-Level Meeting on Global AI Governance, which opened in Shanghai on 17 July and runs until 20 July.

At the opening ceremony, President Xi Jinping called for a global AI governance system centred on openness, inclusion, safety and shared development.

The conference brings together government representatives, international organisations, researchers and technology companies to discuss AI development and governance.

China is using the event to present international cooperation, open-source development and capacity-building for developing countries as central elements of its approach to global AI governance.

Why does it matter?

China’s position reflects growing competition over who shapes global AI standards, infrastructure and access. Its emphasis on open-source models and capacity-building may appeal to developing countries seeking affordable technology, while also strengthening Beijing’s influence in global AI governance. The debate, therefore, concerns both narrowing the AI divide and the strategic dependencies that may emerge around models, infrastructure and technical standards.

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

Australia announces national AI standards and new AI office

Australia has announced plans for mandatory national AI standards and established a new Office of AI within the Department of the Prime Minister and Cabinet.

Prime Minister Anthony Albanese said the measures would coordinate AI policy across government, protect national interests and strengthen Australia’s sovereign technological capacity.

The proposed framework would bring large AI data centres under nationally consistent rules covering their location, infrastructure, energy use and water consumption.

Albanese plans to seek agreement from state and territory leaders at the National Cabinet in August 2026, with legislation expected to be introduced to Parliament in early 2027.

The next generation of large data centres could be legally required to underwrite new power supplies, cover their share of grid connection costs and add at least as much electricity to the grid as they consume.

Facilities would also be required to minimise water use, improve energy efficiency and fund any additional water infrastructure they need.

The government plans to strengthen protections for Australian writers, musicians, artists and journalists, arguing that creators should retain control over whether and on what terms their work is used to train AI systems.

The new Office of AI will coordinate government work across industry, employment, education, energy, copyright, online safety, productivity and national security.

Australia also wants to attract frontier AI investment and increase domestic development, linking sovereign AI capability to secure employment, economic resilience and reduced dependence on foreign technology platforms.

Why does it matter?

Australia’s announcement connects AI governance with physical infrastructure, copyright, energy policy and technological sovereignty. National rules could reduce fragmented approval processes while requiring large data centres to account for their effects on electricity, water and local infrastructure. Central coordination through the Office of AI may also help align policies that currently span several departments. However, the eventual impact will depend on the legislation, enforcement mechanisms and copyright protections that the government develops.

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

UN chief warns against concentrated AI power

UN Secretary-General António Guterres has called for AI to be governed through broad international cooperation rather than shaped by a small group of governments and technology companies. Speaking at the opening of the World Artificial Intelligence Conference in Shanghai, he described AI as “humanity’s greatest opportunity in the 21st century”, while warning that it could also become “one of its greatest risks”.

“Technology that will shape the future of humanity must be shaped by all of humanity”, Guterres said. He stressed that AI governance cannot be governed by a handful of countries or companies” and that “every nation needs a seat at the table”. Recent UN efforts include the Global Digital Compact, the Independent International Scientific Panel on AI and the first Global Dialogue on AI Governance, held in Geneva earlier in July.

AI could support medical research, education, food systems, employment and progress towards the sustainable development goals, but unequal access remains a major concern. “One-third of humanity is still offline,” Guterres noted, warning that concentrated computing capacity, expertise and investment could produce “greater inequalities, greater divides in income, in opportunity, in security”. More than 20 countries have nominated centres for a UN-supported capacity-building network, while recommendations for a Global Fund for AI are also expected.

The UN chief identified capacity development, common safety standards and environmental sustainability as central priorities. “Human rights must be protected”, the UN chief said. “Humans must keep control over every life-and-death decision.” He also stressed that “no AI system should be put in a child’s hands before it has been proven safe”.

Guterres called on major AI companies to disclose their environmental impact and use renewable energy by 2030. “The defining question is whether that transformation will reduce inequalities or reinforce them,” he said. “Whether it will concentrate power or expand opportunity.”

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

EU adopts AI governance principles for health technology assessments

The EU’s Health Technology Assessment Coordination Group has adopted principles governing the use of AI in health technology assessments, requiring transparency, human oversight and accountability when AI assists in preparing evidence dossiers.

The guidance apply to dossiers submitted by health technology developers for Joint Clinical Assessments under the EU Health Technology Assessment Regulation. AI may be used to identify studies, extract data, assess the risk of bias, analyse evidence and prepare reports.

The group recognises that AI could improve the efficiency of evidence preparation but warns that inadequate oversight may compromise completeness, methodological quality and scientific rigour.

Health technology developers will remain fully responsible for the content, methods and conclusions of submitted dossiers, including decisions on whether AI is used, how it contributes to evidence synthesis and whether its outputs are scientifically valid.

Human oversight must be maintained throughout every AI-assisted stage, and no part of the assessment process should be fully automated without a clearly accountable individual.

Any AI-assisted step must be disclosed in the dossier, including information retrieval, study screening, data extraction, risk-of-bias assessment, evidence analysis and report preparation. Developers must also identify the AI tools used, including their name, version, developer and intended purpose.

Where commercial tools have been adapted, those modifications should be described, while prompts must be retained and made available during assessments on request.

The principles also require AI use to comply with applicable copyright, data protection and EU AI Act requirements, particularly given that submitted dossiers may later be published.

Why does it matter?

AI could accelerate the preparation and review of clinical evidence, but undisclosed automation or inaccurate outputs could influence decisions on which medicines and health technologies reach patients. By requiring human accountability, methodological transparency and disclosure of AI tools and prompts, the EU is seeking to preserve the scientific integrity of health technology assessments.

The principles also represent an early example of sector-specific AI governance under the EU’s broader regulatory framework. Rather than restricting AI use, they establish conditions for trustworthy deployment, illustrating how transparency and human oversight are becoming central requirements for AI-assisted decision-making in highly regulated fields.

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

Report urges Hong Kong to strengthen AI adoption strategy

Hong Kong should draw on the experiences of Singapore and the United Kingdom to accelerate AI adoption, according to research published by the Legislative Council Secretariat.

Although Hong Kong has identified AI as a priority industry, it has yet to adopt a standalone economy-wide AI strategy. A 2025 survey found that only around 2% of local organisations were fully prepared for AI, compared with a global average of 13%.

Hong Kong ranked 20th in the International Monetary Fund’s AI Preparedness Index, behind Singapore, Japan, South Korea and the UK. While it scored highly for digital infrastructure and innovation, it ranked considerably lower in human capital, labour policy, regulation and ethics.

The report identifies fragmented governance, limited in-house expertise and insufficient computing capacity as major barriers. Around 44% of Hong Kong businesses reportedly struggled to secure adequate processing power, while many, particularly smaller firms, lacked support for redesigning workflows and integrating legacy data systems.

Singapore has addressed these barriers through coordinated funding, diagnostics and technical assistance. Grants subsidise approved AI tools, while government programmes connect businesses with consultants, engineers and major cloud providers. By October 2025, its Enterprise Compute Initiative had linked around 1,000 companies with cloud partners.

However, the report notes that Singapore also faces challenges in moving from experimentation to widespread organisational transformation. AI adoption reached 62.5% among larger companies and 14.5% among SMEs in 2024, yet only 4% of firms had integrated AI into their core business processes by 2026.

The UK has concentrated support on SMEs and traditionally less digitised sectors through programmes such as BridgeAI and Made Smarter, which combine funding with mentoring, diagnostics, training and technical collaboration. Business use of AI increased from 9.4% in September 2023 to 25.9% in March 2026, with particularly strong growth in construction, transport and manufacturing.

Both Singapore and the UK have also expanded computing infrastructure. Singapore combines cloud credits with technical support, while the UK is investing in public supercomputers and regional AI Growth Zones, which had attracted £28.2 billion in planned investment by January 2026.

The report concludes that accelerating AI adoption will require more than financial incentives. A coordinated approach combining affordable computing, technical support, workforce development and assistance with business transformation could help organisations move from isolated AI pilots to widespread operational use.

Why does it matter?

The report highlights that digital infrastructure alone is not enough to drive AI adoption. Countries that combine funding with technical assistance, workforce development, computing capacity and coordinated governance appear better positioned to translate AI investment into measurable productivity gains.

For Hong Kong, the findings suggest that competitiveness will depend less on access to AI technologies than on the ability of businesses, particularly SMEs, to integrate them into everyday operations. The report therefore reinforces a broader international trend towards AI adoption policies that emphasise implementation alongside innovation.

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

ONS reports growing concern about AI in Great Britain

Public concern about AI is growing across Great Britain, with more adults believing its risks outweigh its benefits, according to new data from the Office for National Statistics (ONS).

The survey found that 38% of adults believed AI’s risks outweighed its benefits, up from 25% in August 2024. Only 13% said the benefits outweighed the risks, while 43% considered them broadly balanced.

Despite growing concerns, 36% of respondents said AI would benefit them personally, although 27% disagreed, the highest share recorded since the ONS began asking the question in November 2023. Younger adults remained considerably more optimistic than older respondents.

Misinformation, privacy and security emerged as the public’s main concerns. Around 81% of respondents believed AI would make fake information harder to identify, 77% worried personal data could be used without consent and 63% expected greater exposure to cybercrime.

Nearly half of adults under 50 also believed AI could threaten their jobs, while trust remained very low for high-impact uses such as government decision-making (4%) and caregiving (5%).

Public concern contrasted with more limited expectations of personal benefit. While respondents identified education, workplace assistance and household support as AI’s main advantages, 41% said the technology would have no positive impact on their own lives.

Why does it matter?

The findings suggest that public acceptance of AI is becoming a key governance challenge alongside technological development. Growing concerns about misinformation, privacy, cybersecurity and employment could make citizens less willing to embrace AI unless governments and companies demonstrate that effective safeguards are in place.

The survey also highlights a widening gap between rapid AI deployment and public confidence. As AI becomes more deeply integrated into public services and everyday life, trust, transparency and accountability may prove just as important as technical capability in determining how quickly the technology is adopted.

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

OECD urges stronger governance as GenAI transforms higher education

The OECD has called for a more responsible and systematic approach to generative AI (GenAI) in higher education, warning that institutional policies, governance and support have failed to keep pace with the technology’s rapid adoption.

Its latest Education Spotlight finds that GenAI has moved from novelty to near-universal use among students. In the UK, undergraduate use rose from 66% in early 2024 to 95% in 2026, while comparable surveys found adoption above 90% among German students.

Across the EU, an average of 72% of students reported using GenAI during the previous three months in 2025, including 53% for formal education, although adoption ranged from almost 90% in Estonia to just over half in Türkiye.

Academic staff are also adopting GenAI, although at a slower pace. Globally, 61% of academics across 28 countries reported using it weekly, but only 17% considered themselves advanced or expert users. Most use remains focused on routine tasks such as drafting, editing, summarising and preparing teaching materials.

Adoption also varies by discipline, with science, engineering and business generally reporting greater confidence and use than the arts and humanities, while differences remain across gender, socio-economic background and institutional resources.

The diagram on page 4 maps GenAI use across teaching, learning, research, administration, student services and institutional management. Applications range from personalised learning materials and research coding to student support, regulatory compliance and large-scale analysis of internal documents.

The OECD warns that institutional governance has not kept pace with widespread adoption. Many students and staff continue to rely on free consumer AI tools because universities have yet to provide secure institutional alternatives. In the UK, only 38% of institutions actively provided GenAI tools to students in 2026, although this represented a significant increase from 9% two years earlier.

Governance is also lagging. A UNESCO survey found that only 19% of participating institutions had adopted formal AI policies, while a further 42% were still developing them. According to the OECD, this leaves students and staff to make individual decisions on issues such as privacy, copyright, ethics and appropriate academic use.

To support responsible adoption, the OECD identifies five priority policy areas: guidance for responsible use, coordinated compliance and procurement, AI skills development, evidence gathering through evaluation, and support for specialised educational tools.

Examples include national guidance in Australia, Ireland and Finland; shared procurement in France and the Netherlands; staff training in Germany, South Korea and Switzerland; and controlled pilot programmes to evaluate educational outcomes.

The report recommends guaranteeing access to devices, connectivity and secure AI tools, investing in staff development, enforcing privacy and safety standards, supporting research and maintaining meaningful human oversight alongside non-AI alternatives.

Why does it matter?

The report suggests that higher education has entered a new phase in which widespread GenAI use is no longer the primary challenge. Instead, universities and governments must determine how to integrate the technology in ways that strengthen learning while protecting academic integrity, privacy and equity.

The OECD also argues that effective AI adoption requires institutional governance rather than individual experimentation. Secure platforms, clear policies, staff training and evidence-based evaluation will increasingly determine whether GenAI enhances education or simply accelerates existing inequalities and weakens critical thinking.

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