False information on the internet (72% of respondents), the state of the global economy (70%), and terrorism (69%) are considered the main global threats in 25 countries around the world. Concerns about climate change (67%) and the spread of infectious diseases (60%) have declined compared to previous years, especially in high-income countries. The Pew Research Center study also revealed significant demographic differences in terms of age, ideological orientation, income level, and attitude toward the national economy.

Misinformation is the main global concern. The greatest concern is observed in Germany, the Netherlands, Poland, Sweden, the United Kingdom, the United States, and South Korea. Concern has risen particularly sharply in Poland (+20 pp), Sweden (+10) and Hungary (+9) since 2022. This threat is more often cited by people over 50 and supporters of left-wing political views, indicating its link to trust in the media and perceptions of political polarisation.

Terrorism is a major concern in middle-income countries. Terrorist threats are most often a cause for concern in India, Israel, Nigeria, and Turkey. At the same time, the US has seen a 9 percentage point decline in concern since 2020. Middle-income countries show significantly higher levels of concern (79% versus 60% in high-income countries), and concern is particularly pronounced among respondents with right-wing and populist views.

Climate threats, despite the high overall level of attention (67%), are not the dominant concern in any country. In high-income countries, concern has declined since 2022. Liberals and young people are more likely to express concern — for example, in the US, 84% of liberals versus 20% of conservatives consider climate change a serious threat.

Concerns about infectious diseases remain strongest in Argentina, Brazil, and South Africa. However, in high-income countries, this figure has fallen significantly—for example, in Germany from 49% to 28% since 2022. Women in ten countries are more likely than men to cite infections as a serious threat.
The results show that global fears are shifting: misinformation and the economy are coming to the fore, while pandemic threats and climate change are losing relevance, especially in developed countries. However, it is important to consider the limitations of the study: narrow geographical coverage (25 countries), time frame (January - April 2025), cultural differences in the interpretation of questions, and the potential influence of recent events on the answers of the respondents.

The authors of the review show how artificial intelligence can completely transform the quality control system in laboratories where stem cells are grown. Instead of rare manual checks, continuous monitoring with automatic data analysis and forecasts is proposed, which helps to identify problems in advance and adjust the process in real time.

Key quality parameters include:

• Cell shape and health (morphology and viability) — computer vision tracks how cells look and detects deviations in a timely manner.
• Incubator conditions — AI predicts possible failures, such as temperature fluctuations.
• Genetic stability — results from different analyses (so-called multi-omics data) are combined to ensure that cells do not change in undesirable ways.
• Risk of cell culture contamination — algorithms process signals from sensors and warn of possible contamination.

The considered approach utilizes the neural networks for image analysis (accuracy above 90%), prediction algorithms (Random Forest, SVM) for risk assessment, and “digital twins” — computer models that simulate cell culture and help select optimal conditions. An important part is Explainable AI: special visualization methods show why the algorithm made a particular decision, which is important for regulatory approval.
The main barriers are huge amounts of data, lack of standards, and high costs for equipment and staff training. In addition, issues of confidentiality (patient data is protected by laws such as GDPR and HIPAA) and transparency of algorithms are important.

The authors expect that in the future, such systems will become autonomous: laboratories will monitor themselves with almost no human involvement, and data from different centers can be used for joint training of algorithms without transferring personal information.

Researchers have created the Virtual Lab system, which allows AI agents to work in a scientific team format alongside humans. Unlike conventional chatbots, this architecture builds an entire team of virtual “scientists”: each agent performs its own role — biologist, immunologist, machine learning specialist — and the "principal investigator" (PI agent) coordinates the work. There is even a “scientific critic” who checks the team's conclusions for rigor.

The study describes how AI completed a full cycle of developing new nanobodies — miniature antibodies that bind to viral proteins. First, the system selects source molecules (e.g., Ty1 or Nb21), then uses three key tools:

• ESM — a model that predicts which point mutations in the protein will be useful,
• AlphaFold-Multimer — a tool that builds a 3D structure of the protein,
• Rosetta — a program for assessing how strong the binding to the virus will be.

The process is iterative: first, the AI generates a bunch of options, then picks the best ones based on a few metrics, and finally, the researchers check them out in the lab.

As a result, the system created 92 variants of nanobodies, 90% of which “assembled” well in cells, and two mutants showed improved binding to new variants of the coronavirus, including KP.3 and JN.1. For example, the modified nanobody Nb21 not only retained its ability to bind to old strains, but also became effective against new ones.

Virtual Lab demonstrates that AI can be not just a tool, but a full-fledged partner in scientific research. The system helps overcome the complexity of interdisciplinary tasks that usually require a large team of experts. The authors note that AI still has weaknesses: models can be trained on outdated data, AlphaFold's predictions are not always accurate, and the result depends on the quality of the source data and the configuration of queries.

The review analyzes what technologies need to be developed to achieve net zero by mid-century. The authors look at the five largest sources of emissions — industry, transport, heat and buildings, agriculture, and power generation — and assess 18 sub-sectors, as well as three key cross-cutting solutions: carbon capture and storage (CCS), hydrogen, and biomass.

The experts used two metrics:

• TMRL — the level of technology and market readiness, which shows how close a solution is to large-scale implementation (from early laboratory stage to readiness for scaling).
• TCL — the level of confidence that the selected technology will actually perform (low, medium, or high).

'Wildcard' technologies — promising but as yet uncertain technologies that could play a key role — are noted separately.
For most sectors, solutions are already available and ready for implementation (e.g., renewable energy and transport electrification), but heavy industry, aviation, and some agricultural processes will require new breakthroughs—such as low-carbon steel or bioengineering to reduce methane emissions. RD&D (research, development, and demonstration) should focus on early-stage innovation, creating tools for selecting optimal technologies, and supporting the scaling up of mature solutions.

The report emphasizes that technological development alone is not enough: investment in infrastructure, training of specialists, new supply chains, public engagement, and political incentives are also needed.

A study by Stanford's Digital Economy Lab analyzes how the wave of AI adoption has affected employment in 2022–2025. The results are alarming for young professionals: in professions where AI can automate a significant portion of tasks—such as software developers and support staff—employment among workers aged 22–25 has fallen by almost 20%.

Six key findings:

1. Decline in employment among young professionals — especially in “highly automatable” professions where AI replaces routine work.
2. The overall labor market remains stable — employment is growing overall, but growth has stalled for young workers and declined by 6% in vulnerable categories.
3. The professions most affected are those where AI replaces humans rather than assisting them — in areas where AI complements labor (e.g., analytics), employment has remained virtually unchanged.
4. The effect persists even when fluctuations in demand at the level of individual companies and industries are excluded.
5. The adjustment is through employment, not wages — wage levels remain stable, indicating high wage inertia.
6. The results are robust — the effect is observed even when IT professions, remote work, and other alternative explanations are excluded.

The authors emphasize that it is important to ensure that AI complements rather than replaces labor, and to create support programs for those who are entering the labor market.

S&P Global analysts have raised their GDP growth forecast for 2025 for the US, Canada, the eurozone, the UK, and mainland China thanks to unexpectedly strong second-quarter results. However, the forecast for India and Brazil has been lowered due to high US tariffs, which are weighing on trade.

Economists warn that growth may slow in the second half of 2025. Among the reasons are a jump in the effective US tariff rate, the exhaustion of the effect of advance purchases, and restrictive monetary policy. An improvement is expected in 2026, when monetary policy easing and fiscal stimulus measures will take effect.