
A Proposed Framework to Monitor, Safeguard and Govern Autonomous AI-to-AI Communication: A Case Study Inspired by Moltbook | IJCT Volume 13 – Issue 6 | IJCT-V13I6P91
IJCT
International Journal of Computer Techniques
ISSN 2394-2231 · Peer-Reviewed · Open Access
📚 Volume 13, Issue 6
📅 November 16, 2026
📄 Pages 774–780
🔖 ID: IJCT-V13I6P91
Table of Contents
ToggleA Proposed Framework to Monitor, Safeguard and Govern Autonomous AI-to-AI Communication: A Case Study Inspired by Moltbook
Author(s)
Snehal Khode, Snehal Lonkar, Priyanka Khairaw, Nandini Pachauri
Abstract
Autonomous AI agents increasingly exchange messages with one another without a human in the loop, and Moltbook, an agent-only social network launched in early 2026, is one of the first large public examples. This paper reviews fifteen recent studies on agent social networks, multi-agent risks, prompt injection and agent governance, and finds that existing work measures or conceptualises the problem but rarely implements and tests an oversight mechanism. We therefore propose AgentGuard, a five-layer framework covering identity, communication control, content monitoring, policy enforcement and tamper-evident auditing, and evaluate a prototype on 44,376 annotated Moltbook posts and a separate April 2026 export with real author identifiers. Learned detectors outperformed keyword baselines and hand-written rules, and combining rate control with content monitoring gave the best end-to-end result. Human review found 74.0% of injection flags correct but only 4.0% of secret-leak flags, and a spam detector trained on early data failed on later data. Layered monitoring is feasible but needs human labelled data, drift monitoring and stronger injection detection.
Keywords
AI agents; AI-to-AI communication; Moltbook; prompt injection; AI governance; multi agent monitoring
Conclusion
AI-to-AI communication is now a real setting, not a thought experiment. The review shows that Moltbook studies, security studies and governance proposals have developed separately and that no reviewed work evaluates an oversight pipeline on real traffic. We proposed AgentGuard and tested a prototype on real Moltbook data. The main benefits are layered coverage, accountability and a transparent baseline; the main challenges are label noise, false alarms, evasion, drift and privacy. Learned detectors outperformed rules, rate control and content monitoring were complementary, a real injection campaign of 1,735 posts was found and confirmed by human review, and secret-leak rules need redesign. Future work should focus on human-labelled benchmarks, semantic injection detection and continual validation of detectors.
References
[1] Anthropic, "Introducing the Model Context Protocol," Nov. 25, 2024. [Online]. Available: https://www.anthropic.com/news/model-context-protocol
[2] Google, "Announcing the Agent2Agent Protocol (A2A)," Google Developers Blog, Apr. 9, 2025. [Online]. Available: https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/
[3] T. Guo et al., "Large language model based multi-agents: A survey of progress and challenges," in Proc. 33rd Int. Joint Conf. Artif. Intell. (IJCAI-24), 2024, pp. 8048-8057, doi: 10.24963/ijcai.2024/890.
[4] L. Wang et al., "A survey on large language model based autonomous agents," Front. Comput. Sci., vol. 18, no. 6, Art. no. 186345, 2024, doi: 10.1007/s11704-024-40231-1.
[5] F. Perez and I. Ribeiro, "Ignore previous prompt: Attack techniques for language models," arXiv:2211.09527, 2022.
[6] K. Greshake, S. Abdelnabi, S. Mishra, C. Endres, T. Holz, and M. Fritz, "Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection," in Proc. 16th ACM Workshop Artif. Intell. Secur. (AISec), 2023, pp. 79-90, arXiv:2302.12173.
[7] D. Lee and M. Tiwari, "Prompt infection: LLM-to-LLM prompt injection within multi-agent systems," arXiv:2410.07283, 2024.
[8] Y. Jiang, Y. Zhang, X. Shen, M. Backes, and Y. Zhang, "’Humans welcome to observe’: A first look at the agent social network Moltbook," arXiv:2602.10127, 2026.
[9] National Institute of Standards and Technology, "Artificial Intelligence Risk Management Framework (AI RMF 1.0)," NIST, Gaithersburg, MD, USA, Rep. NIST AI 100-1, 2023, doi: 10.6028/NIST.AI.100-1.
[10] European Parliament and Council of the European Union, "Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)," Official J. Eur. Union, 2024.
[11] W. Hou and Z. Ji, "Structural divergence between AI-agent and human social networks in Moltbook," arXiv:2602.15064, 2026.
[12] M. Li, X. Li, and T. Zhou, "Does socialization emerge in AI agent society? A case study of Moltbook," arXiv:2602.14299, 2026.
[13] I.-H. Ting, K. Minetaki, D. Liberona, and M.-E. Wu, "Exploring agent interactions in MoltBook through social network analysis," arXiv:2605.27349, 2026.
[14] W. Brach, "The Moltbook Files: A harmless slopocalypse or humanity’s last experiment," arXiv:2605.07462, 2026.
[15] L. Hammond et al., "Multi-agent risks from advanced AI," Cooperative AI Foundation, Tech. Rep. 1, 2025, arXiv:2502.14143. in
[16] Y. Liu, Y. Jia, R. Geng, J. Jia, and N. Z. Gong, "Formalizing and benchmarking prompt injection attacks and defenses," Proc. 33rd USENIX Secur. Symp., https://www.usenix.org/conference/usenixsecurity24/presentation/liu-yupei 2024. [Online]. Available:
[17] Q. Zhan, Z. Liang, Z. Ying, and D. Kang, "InjecAgent: Benchmarking indirect prompt injections in tool-integrated large language model agents," in Findings Assoc. Comput. Linguistics: ACL 2024, 2024, pp. 10471-10506, doi: 10.18653/v1/2024.findings-acl.624. 6
[18] S. R. Motwani et al., "Secret collusion among AI agents: Multi-agent deception via steganography," in Adv. Neural Inf. Process. Syst. 37 (NeurIPS), 2024, doi: 10.52202/079017-2336.
[19] A. Chan et al., "Visibility into AI agents," in Proc. ACM Conf. Fairness, Accountability, Transparency (FAccT), 2024, arXiv:2401.13138.
[20] A. Chan et al., "Infrastructure for AI agents," arXiv:2501.10114, 2025.
[21] N. Kolt, "Governing AI agents," arXiv:2501.07913, 2025.
[22] M. Cemri et al., "Why do multi-agent LLM systems fail?" arXiv:2503.13657, 2025.
[23] OWASP Foundation, "OWASP Top 10 for LLM applications 2025," 2025. [Online]. Available: https://genai.owasp.org/resource/owasp-top-10-for-llm-applications-2025/
[24] TrustAIRLab, "Moltbook dataset with annotations," Hugging Face, 2026. [Data set]. Available: https://huggingface.co/datasets/TrustAIRLab/Moltbook (Zenodo DOI: 10.5281/zenodo.18512310)
[25] F. Pedregosa et al., "Scikit-learn: Machine learning in Python," J. Mach. Learn. Res., vol. 12, pp. 2825-2830, 2011.
[26] timothyngtinhang, "Moltbook posts and comments, April 1-10, 2026," Kaggle, 2026. [Data set]. Available: https://www.kaggle.com/datasets/timothyngtinhang/moltbook-posts-and-comments-april-110-2026
[27] M. Sokolova and G. Lapalme, "A systematic analysis of performance measures for classification tasks," Inf. Process. Manage., vol. 45, no. 4, pp. 427-437, 2009.
[28] B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap. New York, NY, USA: Chapman & Hall, 1993.
[29] J. Cohen, "A coefficient of agreement for nominal scales," Educ. Psychol. Meas., vol. 20, no. 1, pp. 37-46, 1960
[2] Google, "Announcing the Agent2Agent Protocol (A2A)," Google Developers Blog, Apr. 9, 2025. [Online]. Available: https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/
[3] T. Guo et al., "Large language model based multi-agents: A survey of progress and challenges," in Proc. 33rd Int. Joint Conf. Artif. Intell. (IJCAI-24), 2024, pp. 8048-8057, doi: 10.24963/ijcai.2024/890.
[4] L. Wang et al., "A survey on large language model based autonomous agents," Front. Comput. Sci., vol. 18, no. 6, Art. no. 186345, 2024, doi: 10.1007/s11704-024-40231-1.
[5] F. Perez and I. Ribeiro, "Ignore previous prompt: Attack techniques for language models," arXiv:2211.09527, 2022.
[6] K. Greshake, S. Abdelnabi, S. Mishra, C. Endres, T. Holz, and M. Fritz, "Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection," in Proc. 16th ACM Workshop Artif. Intell. Secur. (AISec), 2023, pp. 79-90, arXiv:2302.12173.
[7] D. Lee and M. Tiwari, "Prompt infection: LLM-to-LLM prompt injection within multi-agent systems," arXiv:2410.07283, 2024.
[8] Y. Jiang, Y. Zhang, X. Shen, M. Backes, and Y. Zhang, "’Humans welcome to observe’: A first look at the agent social network Moltbook," arXiv:2602.10127, 2026.
[9] National Institute of Standards and Technology, "Artificial Intelligence Risk Management Framework (AI RMF 1.0)," NIST, Gaithersburg, MD, USA, Rep. NIST AI 100-1, 2023, doi: 10.6028/NIST.AI.100-1.
[10] European Parliament and Council of the European Union, "Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)," Official J. Eur. Union, 2024.
[11] W. Hou and Z. Ji, "Structural divergence between AI-agent and human social networks in Moltbook," arXiv:2602.15064, 2026.
[12] M. Li, X. Li, and T. Zhou, "Does socialization emerge in AI agent society? A case study of Moltbook," arXiv:2602.14299, 2026.
[13] I.-H. Ting, K. Minetaki, D. Liberona, and M.-E. Wu, "Exploring agent interactions in MoltBook through social network analysis," arXiv:2605.27349, 2026.
[14] W. Brach, "The Moltbook Files: A harmless slopocalypse or humanity’s last experiment," arXiv:2605.07462, 2026.
[15] L. Hammond et al., "Multi-agent risks from advanced AI," Cooperative AI Foundation, Tech. Rep. 1, 2025, arXiv:2502.14143. in
[16] Y. Liu, Y. Jia, R. Geng, J. Jia, and N. Z. Gong, "Formalizing and benchmarking prompt injection attacks and defenses," Proc. 33rd USENIX Secur. Symp., https://www.usenix.org/conference/usenixsecurity24/presentation/liu-yupei 2024. [Online]. Available:
[17] Q. Zhan, Z. Liang, Z. Ying, and D. Kang, "InjecAgent: Benchmarking indirect prompt injections in tool-integrated large language model agents," in Findings Assoc. Comput. Linguistics: ACL 2024, 2024, pp. 10471-10506, doi: 10.18653/v1/2024.findings-acl.624. 6
[18] S. R. Motwani et al., "Secret collusion among AI agents: Multi-agent deception via steganography," in Adv. Neural Inf. Process. Syst. 37 (NeurIPS), 2024, doi: 10.52202/079017-2336.
[19] A. Chan et al., "Visibility into AI agents," in Proc. ACM Conf. Fairness, Accountability, Transparency (FAccT), 2024, arXiv:2401.13138.
[20] A. Chan et al., "Infrastructure for AI agents," arXiv:2501.10114, 2025.
[21] N. Kolt, "Governing AI agents," arXiv:2501.07913, 2025.
[22] M. Cemri et al., "Why do multi-agent LLM systems fail?" arXiv:2503.13657, 2025.
[23] OWASP Foundation, "OWASP Top 10 for LLM applications 2025," 2025. [Online]. Available: https://genai.owasp.org/resource/owasp-top-10-for-llm-applications-2025/
[24] TrustAIRLab, "Moltbook dataset with annotations," Hugging Face, 2026. [Data set]. Available: https://huggingface.co/datasets/TrustAIRLab/Moltbook (Zenodo DOI: 10.5281/zenodo.18512310)
[25] F. Pedregosa et al., "Scikit-learn: Machine learning in Python," J. Mach. Learn. Res., vol. 12, pp. 2825-2830, 2011.
[26] timothyngtinhang, "Moltbook posts and comments, April 1-10, 2026," Kaggle, 2026. [Data set]. Available: https://www.kaggle.com/datasets/timothyngtinhang/moltbook-posts-and-comments-april-110-2026
[27] M. Sokolova and G. Lapalme, "A systematic analysis of performance measures for classification tasks," Inf. Process. Manage., vol. 45, no. 4, pp. 427-437, 2009.
[28] B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap. New York, NY, USA: Chapman & Hall, 1993.
[29] J. Cohen, "A coefficient of agreement for nominal scales," Educ. Psychol. Meas., vol. 20, no. 1, pp. 37-46, 1960
📋 How to Cite This Paper
Snehal Khode, Snehal Lonkar, Priyanka Khairaw, Nandini Pachauri (2026). A Proposed Framework to Monitor, Safeguard and Govern Autonomous AI-to-AI Communication: A Case Study Inspired by Moltbook. International Journal of Computer Techniques, 13(6), 774–780. ISSN: 2394-2231. DOI: https://doi.org/10.5281/zenodo.24285999









