About AICrypt
The intersection of artificial intelligence (AI) and security is attracting growing attention, driven by the need to build secure AI-enabled systems. Cryptography is a central example of this trend, showcasing how AI can both strengthen and challenge existing security mechanisms. On one hand, AI techniques are increasingly used within cryptography itself, for instance, to advance cryptanalysis, improve the effectiveness of physical attacks, and design more robust countermeasures. On the other hand, cryptographic methods are now being explored to address security and privacy issues in AI systems.
As attacks on AI become more sophisticated, there is a pressing need to understand how cryptographic tools can mitigate these threats. This includes, for example, cryptographic backdoors in neural networks, cryptographic techniques for watermarking the outputs of large language models (LLMs), and model-stealing attacks that leverage cryptanalytic methods. These developments highlight a rich, bidirectional relationship between AI and cryptography.
The goal of AICrypt 2026 is to bring together experts from academia and industry, each contributing distinct perspectives on AI and cryptography, to foster knowledge exchange and collaborative innovation. We are particularly interested in:
- - transferring techniques across different cryptographic and AI security applications;
- - developing and strengthening security mechanisms for AI systems; and
- - understanding new threat models and corresponding defenses
We will also review recent advances in this rapidly evolving area, including those presented at previous AICrypt events, to provide participants with a comprehensive view of the current state of the field and open research directions.
Topics of Interest
Authors interested in giving a contributed talk in AICrypt are invited to submit an extended abstract of at most 2 pages (excluding references) on Easychair.
Topics of interest for this workshop include, but are not limited to:
- - Machine learning-based cryptanalysis
- - AI-assisted side-channel and fault attacks
- - Cryptanalysis of public-key and symmetric-key crypto algorithms
- - AI-driven design and evaluation of countermeasures
- - Automated discovery of cryptographic weaknesses
- - Privacy-preserving training and inference (e.g., MPC, HE, DP)
- - Cryptographic enforcement of access control and usage policies for models and data
- - Secure model deployment and key management for AI systems
- - Cryptographic primitives tailored to AI workloads
- - Robust watermarking and fingerprinting of models and LLM outputs
- - Model stealing and extraction attacks related to cryptanalytic techniques
- - Backdoors and trojans in neural networks with cryptographic constructions
- - Attacks on federated learning and other distributed AI settings
- - Formal models and proofs for AI security and privacy
- - New cryptographic assumptions motivated by AI settings
- - Transfer of techniques between traditional cryptography and AI security
- - Case studies and empirical evaluations of AI-crypto systems
We welcome novel research results and position paper contributions that clarify the landscape of cryptography and AI security. Work-in-progress and visionary ideas with a clear
technical core are also encouraged.
Submitted abstracts for contributed talks will be reviewed by the workshop chairs for suitability and interest to the AICrypt audience. There are no formal proceedings published in this workshop. Thus, authors can submit extended abstracts related to works submitted or recently published in other venues or work in progress that they plan to submit elsewhere.
