Adversarial prompt and fine-tuning attacks threaten medical large language models
The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcom...
Saved in:
| Published in: | Nature communications Vol. 16; no. 1; pp. 9011 - 10 |
|---|---|
| Main Authors: | , , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
London
Nature Publishing Group UK
09.10.2025
Nature Publishing Group Nature Portfolio |
| Subjects: | |
| ISSN: | 2041-1723, 2041-1723 |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract | The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks–prompt injections with malicious instructions and fine-tuning with poisoned samples–across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings.
Large language models hold significant potential in healthcare settings. This study exposes their vulnerability in medical applications and demonstrates the inadequacy of existing safeguards, highlighting the need for future studies to develop reliable methods for detecting and mitigating these risks. |
|---|---|
| AbstractList | The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks-prompt injections with malicious instructions and fine-tuning with poisoned samples-across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings.The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks-prompt injections with malicious instructions and fine-tuning with poisoned samples-across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings. The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks-prompt injections with malicious instructions and fine-tuning with poisoned samples-across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings. The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks–prompt injections with malicious instructions and fine-tuning with poisoned samples–across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings. Large language models hold significant potential in healthcare settings. This study exposes their vulnerability in medical applications and demonstrates the inadequacy of existing safeguards, highlighting the need for future studies to develop reliable methods for detecting and mitigating these risks. Abstract The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks–prompt injections with malicious instructions and fine-tuning with poisoned samples–across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings. The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks–prompt injections with malicious instructions and fine-tuning with poisoned samples–across three medical tasks: disease prevention, diagnosis, and treatment. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are vulnerable to malicious manipulation across multiple tasks. We discover that while integrating poisoned data does not markedly degrade overall model performance on medical benchmarks, it can lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings.Large language models hold significant potential in healthcare settings. This study exposes their vulnerability in medical applications and demonstrates the inadequacy of existing safeguards, highlighting the need for future studies to develop reliable methods for detecting and mitigating these risks. |
| ArticleNumber | 9011 |
| Author | Huang, Furong Lu, Zhiyong Yang, Yifan Jin, Qiao |
| Author_xml | – sequence: 1 givenname: Yifan orcidid: 0000-0003-4414-9176 surname: Yang fullname: Yang, Yifan organization: National Library of Medicine (NLM), National Institutes of Health (NIH), University of Maryland at College Park, Department of Computer Science – sequence: 2 givenname: Qiao orcidid: 0000-0002-1268-7239 surname: Jin fullname: Jin, Qiao organization: National Library of Medicine (NLM), National Institutes of Health (NIH) – sequence: 3 givenname: Furong surname: Huang fullname: Huang, Furong organization: University of Maryland at College Park, Department of Computer Science – sequence: 4 givenname: Zhiyong orcidid: 0000-0001-9998-916X surname: Lu fullname: Lu, Zhiyong email: zhiyong.lu@nih.gov organization: National Library of Medicine (NLM), National Institutes of Health (NIH) |
| BackLink | https://www.ncbi.nlm.nih.gov/pubmed/41068092$$D View this record in MEDLINE/PubMed |
| BookMark | eNp9kctu1TAQhi1URMuhL8ACRWLDJuDLOJdlVXGpVAkhwdqa2OOQQ-Ic7KRS3x6fk1IQC7ywR9Y3_1z-5-wszIEYeyn4W8FV8y6BgKouudRlBbySpXjCLiQHUYpaqrO_4nN2mdKe56Na0QA8Y-cgeNXwVl6wL1fujmLCOOBYHOI8HZYCgyv8EKhc1jCEvsBlQfsjFcv3SLhQKCZyg838iLGnfId-xRxMs6MxvWBPPY6JLh_eHfv24f3X60_l7eePN9dXt6UF1SwltpzqqhOdACfRK4tOi9bWjURsQIF2mrSsoe281K3ytRXE66ZTHrh3jqsdu9l03Yx7c4jDhPHezDiY08cce4NxGexIRlkpK-0VcSIACx1acCAa3RIhiCprvdm08gZ-rpQWMw3J0phHo3lNRuUWGl21uZEde_0Pup_XGPKkJwqynDxSrx6otcvbemzv9-IzIDfAxjmlSP4REdwcDTabwSYbbE4GG5GT1JaUMhx6in9q_yfrF2Pnpl0 |
| Cites_doi | 10.48550/arXiv.2307.15043 10.48550/arXiv.2411.11525 10.1136/jme-2023-109549 10.1093/bioinformatics/btae075 10.1038/sdata.2016.35 10.48550/arXiv.2404.15777 10.48550/arXiv.2310.15140 10.1109/ICCV51070.2023.00412 10.4174/astr.2023.104.5.269 10.18653/v1/2024.naacl-long.171 10.3389/frai.2023.1169595 10.1093/bib/bbad493 10.48550/arXiv.2312.10997 10.1109/ICCVW54120.2021.00008 10.48550/arXiv.2307.09288 10.1136/bjo-2023-324438 10.2196/53724 10.48550/arXiv.2303.08774 10.48550/arXiv.1909.06146 10.48550/arXiv.2411.18940 10.1038/s41597-023-02814-8 10.48550/arXiv.2306.05499 10.1038/s41467-024-53081-z 10.3390/app11146421 10.18653/v1/W19-5003 10.1007/978-3-031-26553-2_22 |
| ContentType | Journal Article |
| Copyright | This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2025 2025. This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply. This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the "License"). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License. |
| Copyright_xml | – notice: This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2025 – notice: 2025. This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply. – notice: This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the "License"). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License. |
| DBID | C6C AAYXX CITATION CGR CUY CVF ECM EIF NPM 3V. 7QL 7QP 7QR 7SN 7SS 7ST 7T5 7T7 7TM 7TO 7X7 7XB 88E 8AO 8FD 8FE 8FG 8FH 8FI 8FJ 8FK ABUWG AEUYN AFKRA ARAPS AZQEC BBNVY BENPR BGLVJ BHPHI C1K CCPQU COVID DWQXO FR3 FYUFA GHDGH GNUQQ H94 HCIFZ K9. LK8 M0S M1P M7P P5Z P62 P64 PHGZM PHGZT PIMPY PJZUB PKEHL PPXIY PQEST PQGLB PQQKQ PQUKI PRINS RC3 SOI 7X8 DOA |
| DOI | 10.1038/s41467-025-64062-1 |
| DatabaseName | Springer Nature OA Free Journals (WRLC) CrossRef Medline MEDLINE MEDLINE (Ovid) MEDLINE MEDLINE PubMed ProQuest Central (Corporate) Bacteriology Abstracts (Microbiology B) Calcium & Calcified Tissue Abstracts Chemoreception Abstracts Ecology Abstracts Entomology Abstracts (Full archive) Environment Abstracts Immunology Abstracts Industrial and Applied Microbiology Abstracts (Microbiology A) Nucleic Acids Abstracts Oncogenes and Growth Factors Abstracts Health & Medical Collection ProQuest Central (purchase pre-March 2016) Medical Database (Alumni Edition) ProQuest Pharma Collection Technology Research Database ProQuest SciTech Collection ProQuest Technology Collection ProQuest Natural Science Collection ProQuest Hospital Collection Hospital Premium Collection (Alumni Edition) ProQuest Central (Alumni) (purchase pre-March 2016) ProQuest Central (Alumni) ProQuest One Sustainability ProQuest Central UK/Ireland Advanced Technologies & Computer Science Collection ProQuest Central Essentials Biological Science Collection ProQuest Central Technology Collection Natural Science Collection Environmental Sciences and Pollution Management ProQuest One Coronavirus Research Database ProQuest Central Korea Engineering Research Database Proquest Health Research Premium Collection Health Research Premium Collection (Alumni) ProQuest Central Student AIDS and Cancer Research Abstracts SciTech Premium Collection ProQuest Health & Medical Complete (Alumni) Biological Sciences ProQuest Health & Medical Collection Medical Database Biological Science Database Advanced Technologies & Aerospace Database ProQuest Advanced Technologies & Aerospace Collection Biotechnology and BioEngineering Abstracts Proquest Central Premium ProQuest One Academic (New) Publicly Available Content Database ProQuest Health & Medical Research Collection ProQuest One Academic Middle East (New) ProQuest One Health & Nursing ProQuest One Academic Eastern Edition (DO NOT USE) ProQuest One Applied & Life Sciences ProQuest One Academic (retired) ProQuest One Academic UKI Edition ProQuest Central China Genetics Abstracts Environment Abstracts MEDLINE - Academic DOAJ Directory of Open Access Journals |
| DatabaseTitle | CrossRef MEDLINE Medline Complete MEDLINE with Full Text PubMed MEDLINE (Ovid) Publicly Available Content Database ProQuest Central Student Oncogenes and Growth Factors Abstracts ProQuest Advanced Technologies & Aerospace Collection ProQuest Central Essentials Nucleic Acids Abstracts SciTech Premium Collection ProQuest Central China Environmental Sciences and Pollution Management ProQuest One Applied & Life Sciences ProQuest One Sustainability Health Research Premium Collection Natural Science Collection Health & Medical Research Collection Biological Science Collection Chemoreception Abstracts Industrial and Applied Microbiology Abstracts (Microbiology A) ProQuest Central (New) ProQuest Medical Library (Alumni) Advanced Technologies & Aerospace Collection ProQuest Biological Science Collection ProQuest One Academic Eastern Edition Coronavirus Research Database ProQuest Hospital Collection ProQuest Technology Collection Health Research Premium Collection (Alumni) Biological Science Database Ecology Abstracts ProQuest Hospital Collection (Alumni) Biotechnology and BioEngineering Abstracts Entomology Abstracts ProQuest Health & Medical Complete ProQuest One Academic UKI Edition Engineering Research Database ProQuest One Academic Calcium & Calcified Tissue Abstracts ProQuest One Academic (New) Technology Collection Technology Research Database ProQuest One Academic Middle East (New) ProQuest Health & Medical Complete (Alumni) ProQuest Central (Alumni Edition) ProQuest One Community College ProQuest One Health & Nursing ProQuest Natural Science Collection ProQuest Pharma Collection ProQuest Central ProQuest Health & Medical Research Collection Genetics Abstracts Health and Medicine Complete (Alumni Edition) ProQuest Central Korea Bacteriology Abstracts (Microbiology B) AIDS and Cancer Research Abstracts ProQuest SciTech Collection Advanced Technologies & Aerospace Database ProQuest Medical Library Immunology Abstracts Environment Abstracts ProQuest Central (Alumni) MEDLINE - Academic |
| DatabaseTitleList | MEDLINE - Academic MEDLINE Publicly Available Content Database CrossRef |
| Database_xml | – sequence: 1 dbid: DOA name: DOAJ Directory of Open Access Journals url: https://www.doaj.org/ sourceTypes: Open Website – sequence: 2 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 3 dbid: PIMPY name: Publicly Available Content Database url: http://search.proquest.com/publiccontent sourceTypes: Aggregation Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Biology |
| EISSN | 2041-1723 |
| EndPage | 10 |
| ExternalDocumentID | oai_doaj_org_article_3c2265f3e0ee44c4bac4d41859eea416 41068092 10_1038_s41467_025_64062_1 |
| Genre | Journal Article |
| GrantInformation_xml | – fundername: U.S. Department of Health & Human Services | NIH | U.S. National Library of Medicine (NLM) funderid: https://doi.org/10.13039/100000092 |
| GroupedDBID | --- 0R~ 39C 53G 5VS 70F 7X7 88E 8AO 8FE 8FG 8FH 8FI 8FJ AAHBH AAJSJ AASML ABUWG ACGFO ACGFS ACIWK ACMJI ACPRK ADBBV ADFRT ADMLS ADRAZ AENEX AEUYN AFKRA AFRAH AHMBA ALMA_UNASSIGNED_HOLDINGS AMTXH AOIJS ARAPS ASPBG AVWKF AZFZN BBNVY BCNDV BENPR BGLVJ BHPHI BPHCQ BVXVI C6C CCPQU DIK EBLON EBS EE. EMOBN F5P FEDTE FYUFA GROUPED_DOAJ HCIFZ HMCUK HVGLF HYE HZ~ KQ8 LGEZI LK8 LOTEE M1P M7P M~E NADUK NAO NXXTH O9- OK1 P2P P62 PHGZM PHGZT PIMPY PJZUB PPXIY PQGLB PQQKQ PROAC PSQYO RNS RNT RNTTT RPM SNYQT SV3 TSG UKHRP AAYXX AFFHD CITATION CGR CUY CVF ECM EIF NPM 3V. 7QL 7QP 7QR 7SN 7SS 7ST 7T5 7T7 7TM 7TO 7XB 8FD 8FK AZQEC C1K COVID DWQXO FR3 GNUQQ H94 K9. M48 P64 PKEHL PQEST PQUKI PRINS RC3 SOI 7X8 |
| ID | FETCH-LOGICAL-c438t-a90e76b1b14d2af3cad519c782aa84345d5e52749bf2593f7c1e078b3f40fdd03 |
| IEDL.DBID | P5Z |
| ISICitedReferencesCount | 0 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001591591400008&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 2041-1723 |
| IngestDate | Mon Oct 13 19:21:11 EDT 2025 Sat Oct 11 06:37:21 EDT 2025 Sat Oct 11 05:36:48 EDT 2025 Mon Oct 13 01:40:50 EDT 2025 Sat Nov 29 07:20:37 EST 2025 Fri Oct 10 01:14:31 EDT 2025 |
| IsDoiOpenAccess | true |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 1 |
| Language | English |
| License | 2025. This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply. |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c438t-a90e76b1b14d2af3cad519c782aa84345d5e52749bf2593f7c1e078b3f40fdd03 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23 |
| ORCID | 0000-0001-9998-916X 0000-0002-1268-7239 0000-0003-4414-9176 |
| OpenAccessLink | https://www.proquest.com/docview/3259441629?pq-origsite=%requestingapplication% |
| PMID | 41068092 |
| PQID | 3259441629 |
| PQPubID | 546298 |
| PageCount | 10 |
| ParticipantIDs | doaj_primary_oai_doaj_org_article_3c2265f3e0ee44c4bac4d41859eea416 proquest_miscellaneous_3259856959 proquest_journals_3259441629 pubmed_primary_41068092 crossref_primary_10_1038_s41467_025_64062_1 springer_journals_10_1038_s41467_025_64062_1 |
| PublicationCentury | 2000 |
| PublicationDate | 2025-10-09 |
| PublicationDateYYYYMMDD | 2025-10-09 |
| PublicationDate_xml | – month: 10 year: 2025 text: 2025-10-09 day: 09 |
| PublicationDecade | 2020 |
| PublicationPlace | London |
| PublicationPlace_xml | – name: London – name: England |
| PublicationTitle | Nature communications |
| PublicationTitleAbbrev | Nat Commun |
| PublicationTitleAlternate | Nat Commun |
| PublicationYear | 2025 |
| Publisher | Nature Publishing Group UK Nature Publishing Group Nature Portfolio |
| Publisher_xml | – name: Nature Publishing Group UK – name: Nature Publishing Group – name: Nature Portfolio |
| References | M Balas (64062_CR9) 2024; 50 64062_CR20 AEW Johnson (64062_CR18) 2016; 3 S Tian (64062_CR2) 2024; 25 WHK Chiu (64062_CR8) 2024; 26 D Jin (64062_CR21) 2021; 11 64062_CR13 64062_CR35 64062_CR14 64062_CR36 64062_CR11 64062_CR33 64062_CR12 Z Zhao (64062_CR19) 2023; 10 64062_CR34 64062_CR17 64062_CR15 64062_CR37 64062_CR16 64062_CR31 64062_CR10 64062_CR32 64062_CR30 N Oh (64062_CR6) 2023; 104 Q Jin (64062_CR4) 2024; 40 64062_CR24 Q Jin (64062_CR3) 2024; 15 64062_CR25 64062_CR22 64062_CR1 64062_CR23 64062_CR28 T Dave (64062_CR7) 2023; 6 64062_CR29 64062_CR26 64062_CR5 64062_CR27 |
| References_xml | – ident: 64062_CR17 doi: 10.48550/arXiv.2307.15043 – ident: 64062_CR31 doi: 10.48550/arXiv.2411.11525 – ident: 64062_CR27 – ident: 64062_CR23 – volume: 50 start-page: 90 year: 2024 ident: 64062_CR9 publication-title: J. Med. Ethics doi: 10.1136/jme-2023-109549 – volume: 40 year: 2024 ident: 64062_CR4 publication-title: Bioinformatics doi: 10.1093/bioinformatics/btae075 – ident: 64062_CR29 – volume: 3 year: 2016 ident: 64062_CR18 publication-title: Sci. Data doi: 10.1038/sdata.2016.35 – ident: 64062_CR37 – ident: 64062_CR33 – ident: 64062_CR5 doi: 10.48550/arXiv.2404.15777 – ident: 64062_CR14 – ident: 64062_CR16 doi: 10.48550/arXiv.2310.15140 – ident: 64062_CR32 doi: 10.1109/ICCV51070.2023.00412 – volume: 104 start-page: 269 year: 2023 ident: 64062_CR6 publication-title: Ann. Surg. Treat. Res. doi: 10.4174/astr.2023.104.5.269 – ident: 64062_CR15 doi: 10.18653/v1/2024.naacl-long.171 – volume: 6 year: 2023 ident: 64062_CR7 publication-title: Front. Artif. Intell. doi: 10.3389/frai.2023.1169595 – volume: 25 year: 2024 ident: 64062_CR2 publication-title: Brief. Bioinform. doi: 10.1093/bib/bbad493 – ident: 64062_CR11 doi: 10.48550/arXiv.2312.10997 – ident: 64062_CR26 – ident: 64062_CR35 doi: 10.1109/ICCVW54120.2021.00008 – ident: 64062_CR20 doi: 10.48550/arXiv.2307.09288 – ident: 64062_CR30 – ident: 64062_CR10 doi: 10.1136/bjo-2023-324438 – volume: 26 start-page: e53724 year: 2024 ident: 64062_CR8 publication-title: J. Med. Internet Res. doi: 10.2196/53724 – ident: 64062_CR1 doi: 10.48550/arXiv.2303.08774 – ident: 64062_CR28 – ident: 64062_CR22 doi: 10.48550/arXiv.1909.06146 – ident: 64062_CR25 doi: 10.48550/arXiv.2411.18940 – volume: 10 year: 2023 ident: 64062_CR19 publication-title: Sci. Data doi: 10.1038/s41597-023-02814-8 – ident: 64062_CR36 – ident: 64062_CR12 doi: 10.48550/arXiv.2306.05499 – volume: 15 year: 2024 ident: 64062_CR3 publication-title: Nat. Commun. doi: 10.1038/s41467-024-53081-z – ident: 64062_CR13 – volume: 11 start-page: 6421 year: 2021 ident: 64062_CR21 publication-title: Appl. Sci. doi: 10.3390/app11146421 – ident: 64062_CR24 doi: 10.18653/v1/W19-5003 – ident: 64062_CR34 doi: 10.1007/978-3-031-26553-2_22 |
| SSID | ssj0000391844 |
| Score | 2.4845302 |
| Snippet | The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations,... Abstract The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment... |
| SourceID | doaj proquest pubmed crossref springer |
| SourceType | Open Website Aggregation Database Index Database Publisher |
| StartPage | 9011 |
| SubjectTerms | 631/114/1305 692/308 Benchmarks Computer Security Delivery of Health Care Disease prevention Health care Health services Humanities and Social Sciences Humans Immunization Language Large Language Models Medical imaging multidisciplinary Patients Science Science (multidisciplinary) Ultrasonic imaging Vaccines X-rays |
| SummonAdditionalLinks | – databaseName: DOAJ Directory of Open Access Journals dbid: DOA link: http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1Ni9YwEB5kUfAifltdJYI3Ldsmk7Y5qrh4kEVBZW8hH1MU3O6y7bvgv3eS9H1dUfHitQ0hmUkyz2Qm8wA8I9lGNNHXjN66GiNvqUGTq4NXTnuKrskv5D6_64-OhuNj8_4S1VfKCSvlgYvgDlRggKBHRQ0RYkDvAsZUccUQOUYT6fRl1HPJmcpnsDLsuuD6SqZRw8GM-UxI7K0dGzFZt79Yolyw_08o87cIaTY8hzfhxooYxcsy0ltwhabbcK1wSH6_Ax8ypfLs0kIS3O_J2SLcFMXI8LFeNunaQ7hlSU_pxfIlQUSaxEkJz4hvKQ9cbO8sRabFme_Cp8M3H1-_rVeehDqgGpbamYb6zre-xSjdqIKLjMsC237nBlSooybN3qfxIzs7auxDS4wMvBqxGWNs1D3Ym04negBCBtk3PkpklIFylCbo3vmATVBdpA4reL6VmT0r5TBsDmOrwRYJW5awzRK2bQWvklh3LVMp6_yBFWxXBdt_KbiC_a1S7Lq_Zqt4IgzkOmkqeLr7zTsjhTvcRKeb0mbQndHc5n5R5m4k2CbOESMreLHV7s_O_z6hh_9jQo_gukzLMGUhmH3YW8439Biuhovl63z-JK_jHykA9DA priority: 102 providerName: Directory of Open Access Journals |
| Title | Adversarial prompt and fine-tuning attacks threaten medical large language models |
| URI | https://link.springer.com/article/10.1038/s41467-025-64062-1 https://www.ncbi.nlm.nih.gov/pubmed/41068092 https://www.proquest.com/docview/3259441629 https://www.proquest.com/docview/3259856959 https://doaj.org/article/3c2265f3e0ee44c4bac4d41859eea416 |
| Volume | 16 |
| WOSCitedRecordID | wos001591591400008&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVAON databaseName: DOAJ Directory of Open Access Journals customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: DOA dateStart: 20150101 isFulltext: true titleUrlDefault: https://www.doaj.org/ providerName: Directory of Open Access Journals – providerCode: PRVHPJ databaseName: ROAD: Directory of Open Access Scholarly Resources customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: M~E dateStart: 20100101 isFulltext: true titleUrlDefault: https://road.issn.org providerName: ISSN International Centre – providerCode: PRVPQU databaseName: Advanced Technologies & Aerospace Database customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: P5Z dateStart: 20100101 isFulltext: true titleUrlDefault: https://search.proquest.com/hightechjournals providerName: ProQuest – providerCode: PRVPQU databaseName: Biological Science Database customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: M7P dateStart: 20100101 isFulltext: true titleUrlDefault: http://search.proquest.com/biologicalscijournals providerName: ProQuest – providerCode: PRVPQU databaseName: Health & Medical Collection customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: 7X7 dateStart: 20100101 isFulltext: true titleUrlDefault: https://search.proquest.com/healthcomplete providerName: ProQuest – providerCode: PRVPQU databaseName: ProQuest Central customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: BENPR dateStart: 20100101 isFulltext: true titleUrlDefault: https://www.proquest.com/central providerName: ProQuest – providerCode: PRVPQU databaseName: Publicly Available Content Database customDbUrl: eissn: 2041-1723 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000391844 issn: 2041-1723 databaseCode: PIMPY dateStart: 20100101 isFulltext: true titleUrlDefault: http://search.proquest.com/publiccontent providerName: ProQuest |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1Lb9QwEB7RFiQuvEsDZWUkbhA18SOJT4iiViDBKiBACxfLrwASzW43WST-PR4n2QrxuHDJIbES2zNjf5kZzwfwyNPccelMGtBbkXIXTKoSXqfWMC2MdzqLJ-Q-vCrn82qxkPXocOvGtMppTYwLtVta9JEfsYDTw9ZdUPl0dZ4iaxRGV0cKjR3YwyoJaJi1-LT1sWD184rz8axMxqqjjseVATlci7CV0TT_ZT-KZfv_hDV_i5PG7ef0-v92_AZcG4EneTZoyk245NtbcGWgovxxG95EZuZOoz6S8IWzVU9060gTUGjab9B7QnTf44l80n9BpOlbcjZEecg3TCcnk-uTRHad7g68Pz159_xFOtItpJazqk-1zHxZmNzk3FHdMKtdgHc2QAitK864cMKL8BMrTROGxJrS5j4ADMManjXOZWwfdttl6w-AUEvLzDjKA1jhtKHSilIbyzPLCucLnsDjadLVaqiqoWI0nFVqEJEKIlJRRCpP4Bjlsm2JFbHjjeX6sxoNTDEbgKRomM-859xyoy13WJlHeq_D9CdwOIlHjWbaqQvZJPBw-zgYGEZNdOuXm6FNJQopQpu7gzZse8JzpC6RNIEnk3pcvPzvA7r3777ch6sUNRTTFOQh7PbrjX8Al-33_mu3nsFOuSjjtZrB3vHJvH47i56EGeat1rNoAuFJ_fJ1_fEnKicJEA |
| linkProvider | ProQuest |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1Lb9QwEB6VAoIL70eggJHgBFETP5L4gBCvqlWXFUgF7c34FUCi2WWTBfVP8RsZO8lWiMetB66J5cTxNzNfPPZ8AA88zR2XzqTI3oqUOzSpSnidWsO0MN7pLJ6Qez8pp9NqNpNvNuDHeBYmbKscfWJ01G5uwxr5NkOejqG7oPLp4msaVKNCdnWU0Ohhse-PvuMvW_tk7yXO70NKd14dvNhNB1WB1HJWdamWmS8Lk5ucO6prZrVDFmMxUmpdccaFE17gv5o0NT6S1aXNPcZRw2qe1c5lDPs9BafRj5dhC1k5K9drOqHaesX5cDYnY9V2y6MnCpqxBYZOmua_xL8oE_AnbvtbXjaGu52L_9uHugQXBmJNnvWWcBk2fHMFzvZSm0dX4W1Unm51sDeCIzpcdEQ3jtTIstNuFVaHiO66UHGAdJ8Ck_YNOeyzWORL2C5PxqVdEtWD2mvw7kTGcx02m3njbwKhlpaZcZQjGeO0ptKKUhvLM8sK5wuewKNxktWirxqiYrafVaqHhEJIqAgJlSfwPOBg3TJU_I4X5suPanAgilkkyqJmPvOec8uNttyFykPSe43TncDWCAc1uKFWHWMhgfvr2-hAQlZIN36-6ttUopAC29zo0bd-E54HaRZJE3g8wvG4878P6Na_3-UenNs9eD1Rk73p_m04T4N1hC0Zcgs2u-XK34Ez9lv3uV3ejeZF4MNJw_Qnq2NfEw |
| linkToPdf | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1Lb9QwEB6V8hAX3o-FAkaCE0Sb-JHEB4SAsqJqtVokQBUX41daJJpdNllQ_xq_jrGTbIV43HrgmlhJHH8z89kezwfwyNPMcelMguwtT7hDkyqF14k1TAvjnU7jCbkPe8V0Wu7vy9kG_BjOwoS0ysEnRkft5jaskY8Z8nQM3TmV46pPi5htT54vviZBQSrstA5yGh1Edv3xd5y-Nc92tnGsH1M6ef3u1ZukVxhILGdlm2iZ-iI3mcm4o7piVjtkNBajptYlZ1w44QXO26Sp8PWsKmzmMaYaVvG0ci5l-NwzcLbAOWaY-M3Ex_X6Tqi8XnLen9NJWTluePRKQT82xzBKk-yXWBglA_7Ec3_bo42hb3L5f_5pV-BST7jJi85CrsKGr6_B-U6C8_g6vI2K1I0Odkiwd0eLlujakQrZd9KuwqoR0W0bKhGQ9jAwbF-To253i3wJafRkWPIlUVWouQHvT6U_N2Gzntf-NhBqaZEaRzmSNE4rKq0otLE8tSx3PucjeDIMuFp01URUzAJgpergoRAeKsJDZSN4GTCxbhkqgccL8-WB6h2LYhYJtKiYT73n3HKjLXehIpH0XuPQj2BrgIbq3VOjTnAxgofr2-hYwm6Rrv181bUpRS4FtrnVIXH9JTwLki2SjuDpAM2Th_-9Q3f-_S0P4AKiU-3tTHfvwkUaDCVkasgt2GyXK38Pztlv7edmeT9aGoFPp43Sn6ljaAY |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Adversarial+prompt+and+fine-tuning+attacks+threaten+medical+large+language+models&rft.jtitle=Nature+communications&rft.au=Yang%2C+Yifan&rft.au=Jin%2C+Qiao&rft.au=Huang%2C+Furong&rft.au=Lu%2C+Zhiyong&rft.date=2025-10-09&rft.pub=Nature+Publishing+Group&rft.eissn=2041-1723&rft.volume=16&rft.issue=1&rft.spage=9011&rft_id=info:doi/10.1038%2Fs41467-025-64062-1&rft.externalDBID=HAS_PDF_LINK |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2041-1723&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2041-1723&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2041-1723&client=summon |