Paper 2025/2216
AgentCrypt: Advancing Privacy and (Secure) Computation in AI Agent Collaboration
Abstract
As AI agents increasingly operate in real-world, multi-agent environments, ensuring reliable and context-aware privacy in agent communication is critical, especially to comply with evolving regulatory requirements. Traditional access controls are insufficient, as privacy risks in agentic applications often arise after access is granted; agents may use information in ways that compromise privacy, such as messaging humans, sharing context with other agents, making tool calls, persisting data, or generating arbitrary functions of private information. Existing approaches often treat privacy as a binary constraint, whether data is shareable or not, overlooking nuanced, role-specific, and computation-dependent privacy needs essential for regulatory compliance. Agents, including those based on large language models, are inherently probabilistic and heuristic. There is no formal guarantee of how an agent will behave for any query, making them ill-suited for operations critical to security. To address this, we introduce AgentCrypt, a three-tiered framework for fine-grained, secure agent communication that adds a protection layer atop any AI agent platform. AgentCrypt spans unrestricted data exchange (Level 1) to fully encrypted computation using techniques such as homomorphic encryption (Level 3). Unlike much of the recent work, our approach guarantees that the privacy of tagged data is always preserved—even when the underlying AI model makes errors. AgentCrypt ensures privacy across diverse interactions and enables computation on otherwise inaccessible data, overcoming barriers such as data silos. We implemented and tested it with Langgraph and Google ADK, demonstrating versatility across platforms. We also introduce a benchmark dataset simulating privacy-critical tasks at all privacy levels, enabling systematic evaluation and fostering the development of regulatable machine learning systems for secure agent communication and computation.
Note: Updated version with revised privacy-preserving functionality levels and security definitions
Metadata
- Available format(s)
-
PDF
- Category
- Cryptographic protocols
- Publication info
- Preprint.
- Keywords
- secure AIprivacy-preserving multi-agent systems
- Contact author(s)
- harish @ nyu edu
- History
- 2026-05-08: last of 5 revisions
- 2025-12-08: received
- See all versions
- Short URL
- https://ia.cr/2025/2216
- License
-
CC BY
BibTeX
@misc{cryptoeprint:2025/2216,
author = {Harish Karthikeyan and Yue Guo and Leo de Castro and Antigoni Polychroniadou and Leo Ardon and Udari Madhushani Sehwag and Sumitra Ganesh and Manuela Veloso},
title = {{AgentCrypt}: Advancing Privacy and (Secure) Computation in {AI} Agent Collaboration},
howpublished = {Cryptology {ePrint} Archive, Paper 2025/2216},
year = {2025},
url = {https://eprint.iacr.org/2025/2216}
}
