Potential risks from automation of biodiversity credits | Nature Ecology & Evolution
Agentic artificial intelligence ( AI ), spaceborne monitoring and programmable payments may benefit nature finance, but could also pose new risks ...
Nature 117
Covers: HealthcareChatbots & LLMsHallucinations & FabricationRegulation & PolicyGovernance & Risk ManagementEducationEnergy & EnvironmentAI Safety & Existential RiskPrivacy & SurveillanceMilitary & Defense
AI - hallucinated citations should be treated as research misconduct - Nature
Nature 658, 1356 (2026). doi: https://doi.org/10.1038/d41586-026-03407-2. Competing Interests. The author declares no competing interests.
Hybrid retrieval–verification framework for hallucination reduction in medical LLMs - Nature
Hybrid retrieval–verification framework for hallucination reduction in medical LLMs ... Medical AI · Large language models · NLI-based unsupported- ...
Hallucinations & FabricationHealthcareChatbots & LLMsGovernance & Risk Management
Materiality andriskin the age of pervasive AI sensors | Nature Machine Intelligence
Drawing on a growing body of work on AI risks, we define AI risksas concrete harms that can be experienced by individuals or communities or that …
Early warning of complex climateriskwith integratedartificial intelligence| Nature Communications
We highlight the role of AI in developing multi-hazardEWSs that integrate Meteorological and Geospatial foundation models (FMs) for impact prediction …
AI tools are spotting errors in research papers: inside a growing movement - Nature
Theriskwas found to be overhyped – a mathematical error in the underlying research suggested a key chemical exceeded the safe limit when in fact it …
Integratingartificial intelligencein unmanned vehicles: navigating uncertainties,risks, and ... - Nature
...risksin the rapid development of AI. This study analyzes the assessment phases of AI for UVs, technological advancements, uncertainties, and …
AI hallucinationsare a feature of LLM design, not a bug - Nature
But AI confabulations are integral to how these models work. They are a feature, not a bug. LLMs are trained on texts, not truths. Each text bears …
The global landscape of academic guidelines for generative AI and LLMs - Nature
The integration of generativeartificial intelligence(AI) and large language ...AI on education that balance its benefits with potentialrisks …
Artificial intelligencefor modeling and understanding extreme weather and climate events
AI enhancesriskcommunication by enabling message personalization ... Table 2 Challenges,risksand needs in AI for extreme events. Full size …
A deep learning digital biomarker to detect hypertension and stratify cardiovascularriskfrom ...
We externally validated HTN-AI and demonstrated associations between HTN-AI riskand incident CVD in 56,760 adults at Brigham and Women's Hospital …
The urgency of centering safety-net organizations in AI governance | npj Digital Medicine
Without a reorientation of the AI governance agenda, marginalized patients will disproportionately bear therisksof AI in the US healthcare …
Governance & Risk ManagementAI Safety & Existential RiskHealthcare
Non-linear relationship between arteriosclerosis index and diabetesriskin non-obese east ...
Previous studies have shown a positive association between the arteriosclerosis index (AI) and future diabetesrisk …
Movingartificial intelligencedevelopment to deployment in ophthalmology: randomised ... - Nature
Considering the potentialrisksof AI in ophthalmology, rigorous testing in real-world clinical practice must align with the potential impact on …
Generative AI in psychology - Nature
In this issue, we publish a Viewpoint with reflections from six experts on the promises andrisksof this technology. Generativeartificial …
On the caveats of AI autophagy | Nature Machine Intelligence
As web scraping increases, so does theriskof incorporating synthetic data generated by other models. This means models could be trained on data …
Are the Internet and AI affecting our memory? What the science says - Nature
AI hallucination scan't be stopped — but these techniques can limit their damage. In fact, various studies paint a more complicated picture. Some …
Are the Internet and AI affecting our memory? What the science says - Nature
Because writing can help people to think deeply and come up with original insights, say scholars, students who offload these processes to AI risknot …
Scientists flock to DeepSeek: how they're using the blockbuster AI model - Nature
AI hallucination scan't be stopped — but these techniques can limit their damage. Although R1 still fails on many tasks that researchers might want …
Artificial Intelligencein Healthcare and Medical Communication - Nature
However, the review also pointed out significant concerns, including ethical issues, theriskof misinformation, and the need for transparency inAI- …
Artificial Intelligence in Education - Nature
Chatbot: A software application that uses AI to simulate conversation with users, often through text or voice interactions. References. ChatGPT for …
Transforming diagnosis throughartificial intelligence| npj Digital Medicine - Nature
AI-driven algorithms are increasingly used in healthcare settings to support clinicians with diagnosis, treatment, and patient outcome prediction …
Artificial Intelligencein Education - Nature
Learn how Nature Research Intelligence gives you complete, forward-looking and trustworthy research insights to guide your research strategy …
China's cheap, open AI model DeepSeek thrills scientists - Nature
But LLMs are prone to inventing facts, a phenomenon called 'hallucination', and often struggle to reason through problems …
Hallucinations & FabricationChina & GeopoliticsChatbots & LLMs
AI hallucinationscan't be stopped — but these techniques can limit their damage - Nature
When computer scientist Andy Zou researchesartificial intelligence(AI), he often asks a chatbot to suggest background reading and references. But …
Artificial intelligencein drug development | Nature Medicine
AI also predicts prognostic biomarkers for various cancers, delivering preciseriskscores for survival, recurrence and metastasis. Notably, survival …
Preventing unrestricted and unmonitored AI experimentation in healthcare through ... - Nature
The integration of large language models (LLMs) into electronic health records offers potential benefits but raises significant ethical, legal, …
Privacy preserving strategies for electronic health records in the era of large language models
National regulations like GDPR and HIPAA offer protection frameworks, but specific strategies are needed to mitigate risk in generative AI.Risks …
Privacy & SurveillanceChatbots & LLMsHealthcareRegulation & Policy
Daily briefing: Why kids need to take morerisks: science reveals the benefits of adventurous play
So far,AI can predict which single mutations in a virus will be most advantageous, but it's a long way from forecasting combinations of mutations and …
A caution against customized AI in healthcare | npj Digital Medicine - Nature
This article critiques the shift towards personalized AI in healthcare and other high-stakes domains, cautioning that without careful deliberation …