This research investigates AI-managed household portfolios, finding that large language models recommend undiversified portfolios concentrated in high-momentum, large-cap technology stocks. While returns may exceed market benchmarks, they do not provide abnormal returns when adjusted for risk characteristics. Recommendations are primarily driven by media attention rather than investment skill.

Research demonstrates that Large Language Models exhibit purpose-conditioned bias when informed of downstream tasks. Goal-aware prompting leads to in-sample overfitting and inflated performance before knowledge cutoffs. Results indicate that disclosing objectives compromises neutrality, necessitating the separation of measurement and evaluation in AI-assisted workflows to ensure statistical validity.

Global Nature Watch is an AI-powered Earth monitoring platform developed by the World Resources Institute and Google. It provides near-real-time data on land-use changes across various ecosystems to support nature-based risk assessment and environmental reporting.

This report evaluates climate change mitigation and adaptation strategies for data centres using the ClimaTech database. It assesses the effectiveness of decarbonisation across three scopes and analyses resilience measures for flood, storm, heat, and wildfire risks, providing quantitative performance metrics for various technological solutions.

This report examines the intersection of artificial intelligence and sustainability transitions, detailing opportunities and risks across environment, energy, labour, finance, and democracy. It proposes a three-layered governance framework and a phased roadmap for global scientific alignment, interim safety measures, and a binding international framework convention to ensure societal well-being.

This research investigates the demand for explainable AI in high-stakes credit settings. It finds that decision-makers strategically avoid explanations that reveal racial or gender bias to preserve moral wiggle room. Additionally, behavioural biases cause individuals to undervalue explanations even when they complement private information and improve decision accuracy.

This report examines the hidden human costs of content moderation and data labelling in Colombia, Kenya, and the Philippines. It details systemic physical, psychological, and economic harms caused by lead firm power, informal employment, and intense target pressure, while identifying significant barriers to corporate accountability and worker rights.

This paper examines AI data centre electricity demand and its costs for ratepayers and the environment. It finds that Big Tech market capitalisations have risen far above predicted levels since ChatGPT's 2022 launch, while data centres drive up electricity prices and fossil fuel use. Regulatory reforms and innovations are recommended.

This Chatham House paper examines four trends — rising defence and dual-use investment, the growth of 'patriotic tech', the push for sovereign AI, and concerns over an AI valuation bubble — that could multipolarise the global AI race, and offers recommendations for private sector preparedness.

This NBER working paper quantifies the environmental costs of US data centre electricity use in 2025. Using facility-level data for approximately 2,800 data centres, it estimates gross external damages of $24.6 billion from air pollution and greenhouse gas emissions, with Texas and Virginia accounting for 30% of the national total.

This BHRC report examines how top US companies are responding to mounting pressure on human rights standards. It identifies three corporate pathways: active deregulatory lobbying by Big Oil and Big Tech, quiet retreat from human rights commitments, and continued adherence. Survey data from April 2026 reveals significant reductions in human rights staffing and budgets.

The 2025 RDR Index Big Tech Edition ranks 14 major technology companies on governance, freedom of expression, and privacy. Microsoft leads overall with 50%, followed by Alphabet at 49% and Meta at 47%. Scores are also assessed across eight thematic lenses, including algorithmic transparency, security, and targeted advertising.

This BreakFree Suisse research note examines Swiss institutional investors — including UBS, SNB, Zurich Insurance, and others — holding billions of dollars in US ICE contractors Palantir, AT&T, Geo Group, and CoreCivic. The report argues these investments conflict with the investors' stated human rights policies and ESG commitments.

The Digital Policy Hub, hosted by the Centre for International Governance Innovation (CIGI), is a research and policy platform focused on the governance of transformative technologies. It supports research, analysis and collaboration on topics including artificial intelligence, data governance, digital security, democracy, outer space and environmental impacts of digitalisation.

This report assesses tech firms' responses to regulatory demands for improved online child safety. Despite new duties, children remain exposed to harmful content via personalised feeds. While some platforms committed to enhanced age assurance and grooming protections, further action is required to enforce minimum age policies effectively.

This report examines gender parity in Africa's technology sector. While Africa leads in women's STEM representation globally, significant drop-offs occur at key career stages. The research identifies barriers such as biases and funding gaps, offering actionable strategies to enhance recruitment, support career progression, and increase access to start-up financing.