Atmospheric carbon dioxide reaches a new record

Climate and resilience

Global observations · Implications for long-lived public infrastructure

Global carbon dioxide concentration reaches 426.8 parts per million

Bulletin and announcement: 27 October 2026. Observation year: 2025. Pre-industrial comparison: before 1750.

Global annual mean surface carbon dioxide concentration in 2025: 426.8 parts per million, with uncertainty of plus or minus 0.2 parts per million, 53 per cent above the pre-industrial level. This is atmospheric concentration, not annual emissions.

The World Meteorological Organization’s latest Greenhouse Gas Bulletin reports a globally averaged surface carbon dioxide concentration of 426.8 parts per million in 2025, with an uncertainty of ±0.2 parts per million. That is 53 per cent above the pre-industrial level. Methane and nitrous oxide also reached record concentrations.

Published on 27 October, the bulletin measures gases accumulated in the atmosphere. It does not measure annual emissions. Carbon dioxide rose by 2.8 parts per million between 2024 and 2025: a smaller increase than in the preceding year, while the total concentration continued upwards. The agency links carbon dioxide’s long persistence to warming over decades and centuries.

Health systems · Public finance

Gaborone, Botswana

Botswana brings finance and health ministries together around a 100-day reform plan

United Nations Economic Commission for Africa announcement: 27 October 2026. The account does not give separate dates for the meetings.

Botswana’s agreed health-financing arrangement places the Ministry of Finance as chair and the Ministry of Health as co-chair of a joint technical working group, with a 100-day action plan. This is an agreement, not an announced funding envelope or evidence of improved care.

Botswana and the United Nations Economic Commission for Africa have agreed to establish a joint technical working group on health financing. The Ministry of Finance will chair it, with the Ministry of Health as co-chair and the Commission providing technical support. A 100-day action plan will guide priority measures, according to the 27 October account of engagements in Gaborone.

Discussions covered primary care, national health insurance, procurement and expenditure monitoring. This is an institutional agreement and reform agenda. The announcement specifies no new funding envelope, commencement date for the 100 days or measured improvement in services.

Artificial intelligence · Education and skills

International higher education · Report announced in Singapore

New AI report warns that completed group work can conceal less student discussion

Report announcement: 27 October 2026. The public resource page describes more than 80 global case studies; their individual observation periods are not specified there.

The Digital Education Council and Pearson’s collaborative-learning report reviews more than 80 case studies and maps 15 approaches. These are counts of evidence reviewed and approaches described, not student counts, prevalence rates or measured learning effects.

The Digital Education Council and Pearson announced Designing AI for Collaborative Learning on 27 October. Its public resource page describes an analysis of more than 80 global case studies and maps 15 approaches to using artificial intelligence (AI) in collaborative learning.

The announcement describes students arriving with similar AI-assisted ideas and asking AI to settle disagreements. It also reports students’ preference for tools that organise or explain rather than lead the group’s thinking. These are the publishers’ reported findings. The public material does not establish how prevalent these behaviours are across universities or provide a population-wide causal estimate of lost learning.

Different evidence for different public decisions

Atmospheric observations describe a changing physical environment. Botswana’s agreement describes an administrative route into reform. The education report examines what may be lost inside apparently successful student work. Each exposes a different gap between a visible result and the capability behind it. A building can be completed without sufficient provision for future heat; a reform group can exist before money reaches care; an assignment can be submitted without sustained discussion. These are institutional risks drawn from the three readings, rather than measured outcomes shared by them. For public investment and oversight, the useful connection is the quality of the evidence: what was observed, what was agreed and what remains to be demonstrated.

Sources and evidence

The original publications and their dates, with the scope of the evidence behind each reading.

Climate and resilience

Global surface concentrations observed in 2025. Atmospheric accumulation, rather than annual emissions or a local weather forecast.

World Meteorological Organization: Greenhouse Gas Bulletin No. 2227 October 2026 · observations for 2025Primary observations, uncertainty and pre-industrial comparisons. Atmospheric concentrations, not annual emissions.
World Meteorological Organization: Greenhouse gas levels once again reach record high27 October 2026Explains persistence and annual growth. The bulletin provides the authoritative percentage comparisons.

Health systems · Public finance

An agreement and reform agenda reported on 27 October. No funding envelope, start date for the 100 days or measured service gains announced.

United Nations Economic Commission for Africa: Botswana health-financing cooperation27 October 2026Agreement to establish a working group and pursue a 100-day plan. No new financing envelope or demonstrated service gain announced.

Artificial intelligence · Education and skills

Publishers’ case-study synthesis. Counts describe the material reviewed and approaches mapped, rather than representative prevalence or causal effects.

Digital Education Council and Pearson: Collaborative-learning report announcement27 October 2026Publishers’ account of the findings. The announcement does not establish representative prevalence or causal effect sizes.
Digital Education Council: Designing AI for Collaborative LearningResource accompanying the 27 October announcementThe public summary describes a case-study synthesis. Representative prevalence estimates and causal effect sizes are not supplied there.