Thinkorithm

Weighing Opportunities and Risks

A structured way to think about what AI may make possible, what could go wrong, and how institutions are trying to manage the balance.

Published · 2 min read

Discussions of AI’s future often divide into optimists and pessimists. A more useful approach is to look at specific opportunities and specific risks, and at what is being done about each.

Opportunities with evidence behind them

  • Science. Protein structure prediction, recognised by the 2024 Nobel Prize in Chemistry, has given researchers predicted structures for almost all known proteins.
  • Public services and infrastructure. Machine-learning weather forecasts are now run operationally by ECMWF, which reports a large reduction in the energy needed per forecast.
  • Health care. Regulators have authorised a growing number of AI-enabled medical devices, most often as tools that support clinicians rather than replace them.
  • Everyday productivity. International studies suggest that, for many occupations, AI is more likely to change tasks and augment work than to remove jobs entirely.

Risks that institutions have named

NIST’s profile for generative AI identifies risks that are unique to or made worse by the technology. Among them:

  • Confabulation — confident but false output.
  • Information integrity — easier production of misleading content.
  • Information security — new ways to attack systems, and new attacks on AI systems themselves.
  • Data privacy — exposure of personal information in training data or outputs.
  • Harmful bias — unequal treatment learned from data.
  • Intellectual property — unclear use of protected material.
  • Environmental impacts — the energy and resources needed to train and run large models.

How the balance is being managed

Approaches differ by country, but several are now established. Voluntary frameworks such as the NIST AI Risk Management Framework give organisations a structured way to map, measure and manage risks. International principles from the OECD and UNESCO set shared expectations. The European Union’s AI Act takes a risk-based legal approach: certain practices have been prohibited since February 2025, obligations for providers of general-purpose AI models have applied since August 2025, and requirements for high-risk systems are being phased in.

A way to think about it

Opportunities and risks often come from the same capability: a model that can write convincing text can help a student or mislead a voter. That is why most frameworks focus less on the technology itself and more on how, where and by whom it is used — and on keeping people accountable for the results.

Sources

  1. Nobel Prize Outreach. Press release: The Nobel Prize in Chemistry 2024.
  2. ECMWF. ECMWF’s AI forecasts become operational, 2025.
  3. U.S. Food and Drug Administration. Artificial Intelligence-Enabled Medical Devices.
  4. International Labour Organization. Generative AI and Jobs, Working Paper 96, 2023.
  5. National Institute of Standards and Technology. Generative Artificial Intelligence Profile, NIST AI 600-1, 2024; AI RMF 1.0, NIST AI 100-1, 2023.
  6. European Commission. Timeline for the Implementation of the EU AI Act; Regulation (EU) 2024/1689, EUR-Lex.