_version_ 1866912840538914816
author Seah, Ee Wei
Zheng, Yongsen
Nikshith, Naga
Morsidi, Mahran
Matienzo, Gabriel Waikin Loh
Gay, Nigel
Vij, Akriti
Chua, Benjamin
Ng, En Qi
Johnson, Sharmini
Wilfred, Vanessa
Lee, Wan Sie
Davidson, Anna
Devine, Catherine
Zorer, Erin
Holvey, Gareth
Coppock, Harry
Walpole, James
Wynee, Jerome
Dubois, Magda
Schmatz, Michael
Keane, Patrick
Deverett, Sam
Black, Bill
Yan, Bo
Sabir, Bushra
Sun, Frank
Zhang, Hao
Farlow, Harriet
Zhou, Helen
Dong, Lingming
Lu, Qinghua
Jang, Seung
Abuadbba, Sharif
O'Callaghan, Simon
Ma, Suyu
Howroyd, Tom
Fung, Cyrus
Azadi, Fatemeh
Nejadgholi, Isar
Vishnubhotla, Krishnapriya
Xiong, Pulei
Lohrasbi, Saeedeh
Buffett, Scott
Iqbal, Shahrear
Vajjala, Sowmya
Safont-Andreu, Anna
Massarelli, Luca
van der Wal, Oskar
Möller, Simon
Delaborde, Agnes
Duguépéroux, Joris
Rolin, Nicolas
Gallienne, Romane
Behanzin, Sarah
Seimandi, Tom
Murakami, Akiko
Semitsu, Takayuki
Tsukiji, Teresa
Kinuthia, Angela
Michie, Michael
Kasaon, Stephanie
Wangari, Jean
Baek, Hankyul
Noh, Jaewon
Nam, Kihyuk
Seo, Sang
Shin, Sungpil
Lee, Taewhi
Kim, Yongsu
author_facet Seah, Ee Wei
Zheng, Yongsen
Nikshith, Naga
Morsidi, Mahran
Matienzo, Gabriel Waikin Loh
Gay, Nigel
Vij, Akriti
Chua, Benjamin
Ng, En Qi
Johnson, Sharmini
Wilfred, Vanessa
Lee, Wan Sie
Davidson, Anna
Devine, Catherine
Zorer, Erin
Holvey, Gareth
Coppock, Harry
Walpole, James
Wynee, Jerome
Dubois, Magda
Schmatz, Michael
Keane, Patrick
Deverett, Sam
Black, Bill
Yan, Bo
Sabir, Bushra
Sun, Frank
Zhang, Hao
Farlow, Harriet
Zhou, Helen
Dong, Lingming
Lu, Qinghua
Jang, Seung
Abuadbba, Sharif
O'Callaghan, Simon
Ma, Suyu
Howroyd, Tom
Fung, Cyrus
Azadi, Fatemeh
Nejadgholi, Isar
Vishnubhotla, Krishnapriya
Xiong, Pulei
Lohrasbi, Saeedeh
Buffett, Scott
Iqbal, Shahrear
Vajjala, Sowmya
Safont-Andreu, Anna
Massarelli, Luca
van der Wal, Oskar
Möller, Simon
Delaborde, Agnes
Duguépéroux, Joris
Rolin, Nicolas
Gallienne, Romane
Behanzin, Sarah
Seimandi, Tom
Murakami, Akiko
Semitsu, Takayuki
Tsukiji, Teresa
Kinuthia, Angela
Michie, Michael
Kasaon, Stephanie
Wangari, Jean
Baek, Hankyul
Noh, Jaewon
Nam, Kihyuk
Seo, Sang
Shin, Sungpil
Lee, Taewhi
Kim, Yongsu
contents The rapid rise of autonomous AI systems and advancements in agent capabilities are introducing new risks due to reduced oversight of real-world interactions. Yet agent testing remains nascent and is still a developing science. As AI agents begin to be deployed globally, it is important that they handle different languages and cultures accurately and securely. To address this, participants from The International Network for Advanced AI Measurement, Evaluation and Science, including representatives from Singapore, Japan, Australia, Canada, the European Commission, France, Kenya, South Korea, and the United Kingdom have come together to align approaches to agentic evaluations. This is the third exercise, building on insights from two earlier joint testing exercises conducted by the Network in November 2024 and February 2025. The objective is to further refine best practices for testing advanced AI systems. The exercise was split into two strands: (1) common risks, including leakage of sensitive information and fraud, led by Singapore AISI; and (2) cybersecurity, led by UK AISI. A mix of open and closed-weight models were evaluated against tasks from various public agentic benchmarks. Given the nascency of agentic testing, our primary focus was on understanding methodological issues in conducting such tests, rather than examining test results or model capabilities. This collaboration marks an important step forward as participants work together to advance the science of agentic evaluations.
format Preprint
id arxiv_https___arxiv_org_abs_2601_15679
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Improving Methodologies for Agentic Evaluations Across Domains: Leakage of Sensitive Information, Fraud and Cybersecurity Threats
Seah, Ee Wei
Zheng, Yongsen
Nikshith, Naga
Morsidi, Mahran
Matienzo, Gabriel Waikin Loh
Gay, Nigel
Vij, Akriti
Chua, Benjamin
Ng, En Qi
Johnson, Sharmini
Wilfred, Vanessa
Lee, Wan Sie
Davidson, Anna
Devine, Catherine
Zorer, Erin
Holvey, Gareth
Coppock, Harry
Walpole, James
Wynee, Jerome
Dubois, Magda
Schmatz, Michael
Keane, Patrick
Deverett, Sam
Black, Bill
Yan, Bo
Sabir, Bushra
Sun, Frank
Zhang, Hao
Farlow, Harriet
Zhou, Helen
Dong, Lingming
Lu, Qinghua
Jang, Seung
Abuadbba, Sharif
O'Callaghan, Simon
Ma, Suyu
Howroyd, Tom
Fung, Cyrus
Azadi, Fatemeh
Nejadgholi, Isar
Vishnubhotla, Krishnapriya
Xiong, Pulei
Lohrasbi, Saeedeh
Buffett, Scott
Iqbal, Shahrear
Vajjala, Sowmya
Safont-Andreu, Anna
Massarelli, Luca
van der Wal, Oskar
Möller, Simon
Delaborde, Agnes
Duguépéroux, Joris
Rolin, Nicolas
Gallienne, Romane
Behanzin, Sarah
Seimandi, Tom
Murakami, Akiko
Semitsu, Takayuki
Tsukiji, Teresa
Kinuthia, Angela
Michie, Michael
Kasaon, Stephanie
Wangari, Jean
Baek, Hankyul
Noh, Jaewon
Nam, Kihyuk
Seo, Sang
Shin, Sungpil
Lee, Taewhi
Kim, Yongsu
Artificial Intelligence
The rapid rise of autonomous AI systems and advancements in agent capabilities are introducing new risks due to reduced oversight of real-world interactions. Yet agent testing remains nascent and is still a developing science. As AI agents begin to be deployed globally, it is important that they handle different languages and cultures accurately and securely. To address this, participants from The International Network for Advanced AI Measurement, Evaluation and Science, including representatives from Singapore, Japan, Australia, Canada, the European Commission, France, Kenya, South Korea, and the United Kingdom have come together to align approaches to agentic evaluations. This is the third exercise, building on insights from two earlier joint testing exercises conducted by the Network in November 2024 and February 2025. The objective is to further refine best practices for testing advanced AI systems. The exercise was split into two strands: (1) common risks, including leakage of sensitive information and fraud, led by Singapore AISI; and (2) cybersecurity, led by UK AISI. A mix of open and closed-weight models were evaluated against tasks from various public agentic benchmarks. Given the nascency of agentic testing, our primary focus was on understanding methodological issues in conducting such tests, rather than examining test results or model capabilities. This collaboration marks an important step forward as participants work together to advance the science of agentic evaluations.
title Improving Methodologies for Agentic Evaluations Across Domains: Leakage of Sensitive Information, Fraud and Cybersecurity Threats
topic Artificial Intelligence
url https://arxiv.org/abs/2601.15679