Consumer protection

Anthropic releases new constitution shaping Claude’s AI behaviour

Anthropic has published a new constitution for its AI model Claude, outlining the values, priorities, and behavioural principles designed to guide its development. Released under a Creative Commons licence, the document aims to boost transparency while shaping Claude’s learning and reasoning.

The constitution plays a central role in training, guiding how Claude balances safety, ethics, compliance, and helpfulness. Rather than rigid rules, the framework explains core principles, enabling AI systems to generalise and apply nuanced judgment.

Anthropic says this approach supports more responsible decision-making while improving adaptability.

The updated framework also enables Claude to refine its own training through synthetic data generation and self-evaluation. Using the constitution in training helps future Claude models align behaviour with human values while maintaining safety and oversight.

Anthropic described the constitution as a living document that will evolve alongside AI capabilities. External feedback and ongoing evaluation will guide updates to strengthen alignment, transparency, and responsible AI development.

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

EU urged to accelerate AI deployment under new Apply AI strategy

European policymakers are calling for urgent action to accelerate AI deployment across the EU, particularly among SMEs and scale-ups, as the bloc seeks to strengthen its position in the global AI race.

Backing the European Commission’s Apply AI Strategy, the European Economic and Social Committee said Europe must prioritise trust, reliability, and human-centric design as its core competitive advantages.

The Committee warned that slow implementation, fragmented national approaches, and limited private investment are hampering progress. While the strategy promotes an ‘AI first’ mindset, policymakers stressed the need to balance innovation with strong safeguards for rights and freedoms.

Calls were also made for simpler access to funding, lighter administrative requirements, and stronger regional AI ecosystems. Investment in skills, inclusive governance, and strategic procurement were identified as key pillars for scaling trustworthy AI and strengthening Europe’s digital sovereignty.

Support for frontier AI development was highlighted as essential for reducing reliance on foreign models. Officials argued that building advanced, sovereign AI systems aligned with European values could enable competitive growth across sectors such as healthcare, finance, and industry.

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

GPT-5.2 shows how AI can generate real-world cyber exploits

Advanced language models have demonstrated the ability to generate working exploits for previously unknown software vulnerabilities. Security researcher Sean Heelan tested two systems built on GPT-5.2 and Opus 4.5 by challenging them to exploit a zero-day flaw in the QuickJS JavaScript interpreter.

Across multiple scenarios with varying security protections, GPT-5.2 completed every task, while Opus 4.5 failed only 2. The systems produced more than 40 functional exploits, ranging from basic shell access to complex file-writing operations that bypassed modern defences.

Most challenges were solved in under an hour, with standard attempts costing around $30. Even the most complex exploit, which bypassed protections such as address space layout randomisation, non-executable memory, and seccomp sandboxing, was completed in just over three hours for roughly $50.

The most advanced task required GPT-5.2 to write a specific string to a protected file path without access to operating system functions. The model achieved this by chaining seven function calls through the glibc exit handler mechanism, bypassing shadow stack protections.

The findings suggest exploit development may increasingly depend on computational resources rather than human expertise. While QuickJS is less complex than browsers such as Chrome or Firefox, the approach demonstrated could scale to larger and more secure software environments.

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

YouTube’s 2026 strategy places AI at the heart of moderation and monetisation

As announced yesterday, YouTube is expanding its response to synthetic media by introducing experimental likeness detection tools that allow creators to identify videos where their face appears altered or generated by AI.

The system, modelled conceptually on Content ID, scans newly uploaded videos for visual matches linked to enrolled creators, enabling them to review content and pursue privacy or copyright complaints when misuse is detected.

Participation requires identity verification through government-issued identification and a biometric reference video, positioning facial data as both a protective and governance mechanism.

While the platform stresses consent and limited scope, the approach reflects a broader shift towards biometric enforcement as platforms attempt to manage deepfakes, impersonation, and unauthorised synthetic content at scale.

Alongside likeness detection, YouTube’s 2026 strategy places AI at the centre of content moderation, creator monetisation, and audience experience.

AI tools already shape recommendation systems, content labelling, and automated enforcement, while new features aim to give creators greater control over how their image, voice, and output are reused in synthetic formats.

The move highlights growing tensions between creative empowerment and platform authority, as safeguards against AI misuse increasingly rely on surveillance, verification, and centralised decision-making.

As regulators debate digital identity, biometric data, and synthetic media governance, YouTube’s model signals how private platforms may effectively set standards ahead of formal legislation.

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

Snapchat settles social media addiction lawsuit as landmark trial proceeds

Snapchat’s parent company has settled a social media addiction lawsuit in California just days before the first major trial examining platform harms was set to begin.

The agreement removes Snapchat from one of the three bellwether cases consolidating thousands of claims, while Meta, TikTok and YouTube remain defendants.

These lawsuits mark a legal shift away from debates over user content and towards scrutiny of platform design choices, including recommendation systems and engagement mechanics.

A US judge has already ruled that such features may be responsible for harm, opening the door to liability that section 230 protections may not cover.

Legal observers compare the proceedings to historic litigation against tobacco and opioid companies, warning of substantial damages and regulatory consequences.

A ruling against the remaining platforms could force changes in how social media products are designed, particularly in relation to minors and mental health risks.

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

Why AI systems privilege Western perspectives: ‘The Silicon Gaze’

A new study from the University of Oxford argues that large language models reproduce a distinctly Western hierarchy when asked to evaluate countries, reinforcing long-standing global inequalities through automated judgment.

Analysing more than 20 million English-language responses from ChatGPT’s 4o-mini model, researchers found consistent favouring of wealthy Western nations across subjective comparisons such as intelligence, happiness, creativity, and innovation.

Low-income countries, particularly across Africa, were systematically placed at the bottom of rankings, while Western Europe, the US, and parts of East Asia dominated positive assessments.

According to the study, generative models rely heavily on data availability and dominant narratives, leading to flattened representations that recycle familiar stereotypes instead of reflecting social complexity or cultural diversity.

The researchers describe the phenomenon as the ‘silicon gaze’, a worldview shaped by the priorities of platform owners, developers, and historically uneven training data.

Because large language models are trained on material produced within centuries of structural exclusion, bias emerges not as a malfunction but as an embedded feature of contemporary AI systems.

The findings intensify global debates around AI governance, accountability, and cultural representation, particularly as such systems increasingly influence healthcare, employment screening, education, and public decision-making.

While models are continuously updated, the study underlines the limits of technical mitigation without broader political, regulatory, and epistemic interventions.

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

Hong Kong crypto licensing overhaul draws industry concern

Hong Kong’s proposed crypto licensing overhaul has drawn criticism from industry leaders, who warn it could disrupt compliant firms and deter blockchain exposure.

Under the proposals, the existing allowance enabling firms to allocate up to 10% of fund assets to crypto without additional licensing would be removed. Even minimal exposure would require a full licence, a move the association called disproportionate and harmful to market experimentation.

Concerns also focused on the absence of transitional arrangements. Without a grace period, firms may be forced to suspend operations while licence applications are reviewed.

The association proposed a six- to 12-month transitional window to allow continued activity during regulatory processing.

Further criticism focused on custody rules restricting client assets to SFC-licensed custodians. Industry representatives warned the measure could limit access to early-stage tokens, restrict Web3 investment, and impose unnecessary geographic constraints.

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

Davos 2026 reveals competing visions for AI

AI has dominated debates at Davos 2026, matching traditional concerns such as geopolitics and global trade while prompting deeper reflection on how the technology is reshaping work, governance, and society.

Political leaders, executives, and researchers agreed that AI development has moved beyond experimentation towards widespread implementation.

Microsoft chief executive Satya Nadella argued that AI should deliver tangible benefits for communities and economies, while warning that adoption will remain uneven due to disparities in infrastructure and investment.

Access to energy networks, telecommunications, and capital was identified as a decisive factor in determining which regions can fully deploy advanced systems.

Other voices at Davos 2026 struck a more cautious tone. AI researcher Yoshua Bengio warned against designing systems that appear too human-like, stressing that people may overestimate machine understanding.

Philosopher Yuval Noah Harari echoed those concerns, arguing that societies lack experience in managing human and AI coexistence and should prepare mechanisms to correct failures.

The debate also centred on labour and global competition.

Anthropic’s Dario Amodei highlighted geopolitical risks and predicted disruption to entry-level white-collar jobs. At the same time, Google DeepMind chief Demis Hassabis forecast new forms of employment alongside calls for shared international safety standards.

Together, the discussions underscored growing recognition that AI governance will shape economic and social outcomes for years ahead.

Diplo is live reporting on all sessions from the World Economic Forum 2026 in Davos.

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

UK study tests social media restrictions on children’s mental health

A major UK research project will examine how restricting social media use affects children’s mental health, sleep, and social lives, as governments debate tougher rules for under-16s.

The trial involves around 4,000 pupils from 30 secondary schools in Bradford and represents one of the first large-scale experimental studies of its kind.

Participants aged 12 to 15 will either have their social media use monitored or restricted through a research app limiting access to major platforms to one hour per day and imposing a night-time curfew.

Messaging services such as WhatsApp will remain available instead of being restricted, reflecting their role in family communication.

Researchers from the University of Cambridge and the Bradford Centre for Health Data Science will assess changes in anxiety, depression, sleep patterns, bullying, and time spent with friends and family.

Entire year groups within each school will experience the same conditions to capture social effects across peer networks rather than isolated individuals.

The findings, expected in summer 2027, arrive as UK lawmakers consider proposals for a nationwide ban on social media use by under-16s.

Although independent from government policy debates, the study aims to provide evidence to inform decisions in the UK and other countries weighing similar restrictions.

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

New EU cybersecurity package strengthens resilience and ENISA powers

The European Commission has unveiled a broad cybersecurity package that moves the EU beyond certification reform towards systemic resilience across critical digital infrastructure.

Building on plans to expand EU cybersecurity certification beyond products and services, the revised Cybersecurity Act introduces a risk-based framework for securing ICT supply chains, with particular focus on dependencies, foreign interference, and high-risk third-country suppliers.

A central shift concerns supply-chain security as a geopolitical issue. The proposal enables mandatory derisking of mobile telecommunications networks, reinforcing earlier efforts under the 5G security toolbox.

Certification reform continues through a redesigned European Cybersecurity Certification Framework, promising clearer governance, faster scheme development, and voluntary certification that can cover organisational cyber posture alongside technical compliance.

The package also tackles regulatory complexity. Targeted amendments to the NIS2 Directive aim to ease compliance for tens of thousands of companies by clarifying jurisdictional rules, introducing a new ‘small mid-cap’ category, and streamlining incident reporting through a single EU entry point.

Enhanced ransomware data collection and cross-border supervision are intended to reduce fragmentation while strengthening enforcement consistency.

ENISA’s role is further expanded from coordination towards operational support. The agency would issue early threat alerts, assist in ransomware recovery with national authorities and Europol, and develop EU-wide vulnerability management and skills attestation schemes.

Together, the measures signal a shift from fragmented safeguards towards a more integrated model of European cyber sovereignty.

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