Saved in:
Bibliographic Details
Main Authors: Bengio, Yoshua, Clare, Stephen, Prunkl, Carina, Rismani, Shalaleh, Andriushchenko, Maksym, Bucknall, Ben, Fox, Philip, Hu, Tiancheng, Jones, Cameron, Manning, Sam, Maslej, Nestor, Mavroudis, Vasilios, McGlynn, Conor, Murray, Malcolm, Stix, Charlotte, Velasco, Lucia, Wheeler, Nicole, Privitera, Daniel, Mindermann, Sören, Acemoglu, Daron, Dietterich, Thomas G., Heintz, Fredrik, Hinton, Geoffrey, Jennings, Nick, Leavy, Susan, Ludermir, Teresa, Marda, Vidushi, Margetts, Helen, McDermid, John, Munga, Jane, Narayanan, Arvind, Nelson, Alondra, Neppel, Clara, Ramchurn, Gopal, Russell, Stuart, Schaake, Marietje, Schölkopf, Bernhard, Soto, Alavaro, Tiedrich, Lee, Varoquaux, Gaël, Yao, Andrew, Zhang, Ya-Qin, Aguirre, Leandro, Ajala, Olubunmi, AlMalek, Fahad Albalawi Noora, Busch, Christian, Carvalho, André, Collas, Jonathan, Gill, Amandeep, Hatip, Ahmet, Heikkilä, Juha, Johnson, Chris, Jolly, Gill, Katzir, Ziv, Kerema, Mary, Kitano, Hiroaki, Krüger, Antonio, McLysaght, Aoife, Molchanovskyi, Oleksii, Monti, Andrea, Lee, Kyoung Mu, Nemer, Mona, Oliver, Nuria, Pezoa, Raquel, Plonk, Audrey, Portillo, José, Ravindran, Balaraman, Riza, Hammam, Rugege, Crystal, Sheikh, Haroon, Wong, Denise, Zeng, Yi, Zhu, Liming
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.13653
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909848372772864
author Bengio, Yoshua
Clare, Stephen
Prunkl, Carina
Rismani, Shalaleh
Andriushchenko, Maksym
Bucknall, Ben
Fox, Philip
Hu, Tiancheng
Jones, Cameron
Manning, Sam
Maslej, Nestor
Mavroudis, Vasilios
McGlynn, Conor
Murray, Malcolm
Stix, Charlotte
Velasco, Lucia
Wheeler, Nicole
Privitera, Daniel
Mindermann, Sören
Acemoglu, Daron
Dietterich, Thomas G.
Heintz, Fredrik
Hinton, Geoffrey
Jennings, Nick
Leavy, Susan
Ludermir, Teresa
Marda, Vidushi
Margetts, Helen
McDermid, John
Munga, Jane
Narayanan, Arvind
Nelson, Alondra
Neppel, Clara
Ramchurn, Gopal
Russell, Stuart
Schaake, Marietje
Schölkopf, Bernhard
Soto, Alavaro
Tiedrich, Lee
Varoquaux, Gaël
Yao, Andrew
Zhang, Ya-Qin
Aguirre, Leandro
Ajala, Olubunmi
AlMalek, Fahad Albalawi Noora
Busch, Christian
Carvalho, André
Collas, Jonathan
Gill, Amandeep
Hatip, Ahmet
Heikkilä, Juha
Johnson, Chris
Jolly, Gill
Katzir, Ziv
Kerema, Mary
Kitano, Hiroaki
Krüger, Antonio
McLysaght, Aoife
Molchanovskyi, Oleksii
Monti, Andrea
Lee, Kyoung Mu
Nemer, Mona
Oliver, Nuria
Pezoa, Raquel
Plonk, Audrey
Portillo, José
Ravindran, Balaraman
Riza, Hammam
Rugege, Crystal
Sheikh, Haroon
Wong, Denise
Zeng, Yi
Zhu, Liming
author_facet Bengio, Yoshua
Clare, Stephen
Prunkl, Carina
Rismani, Shalaleh
Andriushchenko, Maksym
Bucknall, Ben
Fox, Philip
Hu, Tiancheng
Jones, Cameron
Manning, Sam
Maslej, Nestor
Mavroudis, Vasilios
McGlynn, Conor
Murray, Malcolm
Stix, Charlotte
Velasco, Lucia
Wheeler, Nicole
Privitera, Daniel
Mindermann, Sören
Acemoglu, Daron
Dietterich, Thomas G.
Heintz, Fredrik
Hinton, Geoffrey
Jennings, Nick
Leavy, Susan
Ludermir, Teresa
Marda, Vidushi
Margetts, Helen
McDermid, John
Munga, Jane
Narayanan, Arvind
Nelson, Alondra
Neppel, Clara
Ramchurn, Gopal
Russell, Stuart
Schaake, Marietje
Schölkopf, Bernhard
Soto, Alavaro
Tiedrich, Lee
Varoquaux, Gaël
Yao, Andrew
Zhang, Ya-Qin
Aguirre, Leandro
Ajala, Olubunmi
AlMalek, Fahad Albalawi Noora
Busch, Christian
Carvalho, André
Collas, Jonathan
Gill, Amandeep
Hatip, Ahmet
Heikkilä, Juha
Johnson, Chris
Jolly, Gill
Katzir, Ziv
Kerema, Mary
Kitano, Hiroaki
Krüger, Antonio
McLysaght, Aoife
Molchanovskyi, Oleksii
Monti, Andrea
Lee, Kyoung Mu
Nemer, Mona
Oliver, Nuria
Pezoa, Raquel
Plonk, Audrey
Portillo, José
Ravindran, Balaraman
Riza, Hammam
Rugege, Crystal
Sheikh, Haroon
Wong, Denise
Zeng, Yi
Zhu, Liming
contents Since the publication of the first International AI Safety Report, AI capabilities have continued to improve across key domains. New training techniques that teach AI systems to reason step-by-step and inference-time enhancements have primarily driven these advances, rather than simply training larger models. As a result, general-purpose AI systems can solve more complex problems in a range of domains, from scientific research to software development. Their performance on benchmarks that measure performance in coding, mathematics, and answering expert-level science questions has continued to improve, though reliability challenges persist, with systems excelling on some tasks while failing completely on others. These capability improvements also have implications for multiple risks, including risks from biological weapons and cyber attacks. Finally, they pose new challenges for monitoring and controllability. This update examines how AI capabilities have improved since the first Report, then focuses on key risk areas where substantial new evidence warrants updated assessments.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13653
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle International AI Safety Report 2025: First Key Update: Capabilities and Risk Implications
Bengio, Yoshua
Clare, Stephen
Prunkl, Carina
Rismani, Shalaleh
Andriushchenko, Maksym
Bucknall, Ben
Fox, Philip
Hu, Tiancheng
Jones, Cameron
Manning, Sam
Maslej, Nestor
Mavroudis, Vasilios
McGlynn, Conor
Murray, Malcolm
Stix, Charlotte
Velasco, Lucia
Wheeler, Nicole
Privitera, Daniel
Mindermann, Sören
Acemoglu, Daron
Dietterich, Thomas G.
Heintz, Fredrik
Hinton, Geoffrey
Jennings, Nick
Leavy, Susan
Ludermir, Teresa
Marda, Vidushi
Margetts, Helen
McDermid, John
Munga, Jane
Narayanan, Arvind
Nelson, Alondra
Neppel, Clara
Ramchurn, Gopal
Russell, Stuart
Schaake, Marietje
Schölkopf, Bernhard
Soto, Alavaro
Tiedrich, Lee
Varoquaux, Gaël
Yao, Andrew
Zhang, Ya-Qin
Aguirre, Leandro
Ajala, Olubunmi
AlMalek, Fahad Albalawi Noora
Busch, Christian
Carvalho, André
Collas, Jonathan
Gill, Amandeep
Hatip, Ahmet
Heikkilä, Juha
Johnson, Chris
Jolly, Gill
Katzir, Ziv
Kerema, Mary
Kitano, Hiroaki
Krüger, Antonio
McLysaght, Aoife
Molchanovskyi, Oleksii
Monti, Andrea
Lee, Kyoung Mu
Nemer, Mona
Oliver, Nuria
Pezoa, Raquel
Plonk, Audrey
Portillo, José
Ravindran, Balaraman
Riza, Hammam
Rugege, Crystal
Sheikh, Haroon
Wong, Denise
Zeng, Yi
Zhu, Liming
Computers and Society
Since the publication of the first International AI Safety Report, AI capabilities have continued to improve across key domains. New training techniques that teach AI systems to reason step-by-step and inference-time enhancements have primarily driven these advances, rather than simply training larger models. As a result, general-purpose AI systems can solve more complex problems in a range of domains, from scientific research to software development. Their performance on benchmarks that measure performance in coding, mathematics, and answering expert-level science questions has continued to improve, though reliability challenges persist, with systems excelling on some tasks while failing completely on others. These capability improvements also have implications for multiple risks, including risks from biological weapons and cyber attacks. Finally, they pose new challenges for monitoring and controllability. This update examines how AI capabilities have improved since the first Report, then focuses on key risk areas where substantial new evidence warrants updated assessments.
title International AI Safety Report 2025: First Key Update: Capabilities and Risk Implications
topic Computers and Society
url https://arxiv.org/abs/2510.13653