Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense

Jiacheng Liu, Yaxin Luo, Jiacheng Cui, Xinyi Shang, Xiaohan Zhao, Zhiqiang Shen
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:77683-77700, 2026.

Abstract

The rapid evolution of GUI-enabled agents has rendered traditional CAPTCHAs obsolete. While previous benchmarks like OpenCaptchaWorld established a baseline for evaluating multimodal agents, recent advancements in reasoning-heavy models, such as Gemini3-Pro-High and GPT-5.2-Xhigh have effectively collapsed this security barrier, achieving pass rates as high as 90% on complex logic puzzles like ”Bingo”. In response, we introduce Next-Gen CAPTCHAs, a scalable defense framework designed to secure the next-generation web against the advanced agents. Unlike static datasets, our benchmark is built upon a robust data generation pipeline, allowing for large-scale and easily scalable evaluations, notably, for backend-supported types, our system is capable of generating effectively unbounded CAPTCHA instances. We exploit the persistent human–agent “Cognitive Gap” in interactive perception, memory, decision-making, and action. By engineering dynamic tasks that require adaptive intuition rather than granular planning, we re-establish a robust distinction between biological users and artificial agents, offering a scalable and diverse defense mechanism for the agentic era.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-liu26cr, title = {Next-Gen {CAPTCHA}s: Leveraging the Cognitive Gap for Scalable and Diverse {GUI}-Agent Defense}, author = {Liu, Jiacheng and Luo, Yaxin and Cui, Jiacheng and Shang, Xinyi and Zhao, Xiaohan and Shen, Zhiqiang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {77683--77700}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/liu26cr/liu26cr.pdf}, url = {https://proceedings.mlr.press/v306/liu26cr.html}, abstract = {The rapid evolution of GUI-enabled agents has rendered traditional CAPTCHAs obsolete. While previous benchmarks like OpenCaptchaWorld established a baseline for evaluating multimodal agents, recent advancements in reasoning-heavy models, such as Gemini3-Pro-High and GPT-5.2-Xhigh have effectively collapsed this security barrier, achieving pass rates as high as 90% on complex logic puzzles like ”Bingo”. In response, we introduce Next-Gen CAPTCHAs, a scalable defense framework designed to secure the next-generation web against the advanced agents. Unlike static datasets, our benchmark is built upon a robust data generation pipeline, allowing for large-scale and easily scalable evaluations, notably, for backend-supported types, our system is capable of generating effectively unbounded CAPTCHA instances. We exploit the persistent human–agent “Cognitive Gap” in interactive perception, memory, decision-making, and action. By engineering dynamic tasks that require adaptive intuition rather than granular planning, we re-establish a robust distinction between biological users and artificial agents, offering a scalable and diverse defense mechanism for the agentic era.} }
Endnote
%0 Conference Paper %T Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense %A Jiacheng Liu %A Yaxin Luo %A Jiacheng Cui %A Xinyi Shang %A Xiaohan Zhao %A Zhiqiang Shen %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-liu26cr %I PMLR %P 77683--77700 %U https://proceedings.mlr.press/v306/liu26cr.html %V 306 %X The rapid evolution of GUI-enabled agents has rendered traditional CAPTCHAs obsolete. While previous benchmarks like OpenCaptchaWorld established a baseline for evaluating multimodal agents, recent advancements in reasoning-heavy models, such as Gemini3-Pro-High and GPT-5.2-Xhigh have effectively collapsed this security barrier, achieving pass rates as high as 90% on complex logic puzzles like ”Bingo”. In response, we introduce Next-Gen CAPTCHAs, a scalable defense framework designed to secure the next-generation web against the advanced agents. Unlike static datasets, our benchmark is built upon a robust data generation pipeline, allowing for large-scale and easily scalable evaluations, notably, for backend-supported types, our system is capable of generating effectively unbounded CAPTCHA instances. We exploit the persistent human–agent “Cognitive Gap” in interactive perception, memory, decision-making, and action. By engineering dynamic tasks that require adaptive intuition rather than granular planning, we re-establish a robust distinction between biological users and artificial agents, offering a scalable and diverse defense mechanism for the agentic era.
APA
Liu, J., Luo, Y., Cui, J., Shang, X., Zhao, X. & Shen, Z.. (2026). Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:77683-77700 Available from https://proceedings.mlr.press/v306/liu26cr.html.

Related Material