In a year defined by both Artificial Intelligence acceleration and rising skepticism about machine intelligence, the ZK/AI Summit held on April 28, 2025 in Dubai came at just the right time, introducing the intersection of AI and verifiable computation. Hosted by House of ZK and @lagrangedev, and supported by sponsors @OpenledgerHQ, @MetisL2, @ProjectZKM, @AEON_Community, and @aztecnetwork, the event brought together the leading projects pioneering the ZK/AI domain to chart a shared vision: a future where intelligent systems are not only powerful, but provably trustworthy. Oversold with 3,000+ registrants, 500+ live attendees, and 200,000+ livestream views, the summit reflected the explosive momentum behind this fascinating new paradigm.
The event opened with a strong keynote from Ismael Hishon-Rezaizadeh (@Ismael_H_R), Founder and CEO of Lagrange, whose remarks set the tone for the day. Zero-knowledge is the key to ensuring AI systems are verifiable at scale. “Without ZKML (zero-knowledge machine learning), we can’t base the next generation of intelligent systems on top of a secure infrastructure of AI,” noted Ismael. Across keynotes and panels, it became clear that ZKML is no longer theoretical, it’s already here reshaping how we build AI systems.
Positioning zero-knowledge proofs as essential to making AI safe, transparent, and infrastructure-ready, Ismael introduced DeepProve - Lagrange’s ZKML system that has become the fastest-performing ZKML library in the industry. With inference benchmarking up to 700x faster than existing solutions, DeepProve is already in use across dozens of projects and is powering use cases in healthcare, content moderation, and on-chain trading agents. Lagrange and its partners are now extending DeepProve’s reach to domains such as national defense against deepfakes and zero-knowledge auditing for AI agents in DeFi. Framing verifiability not as a security add-on but as a new primitive for the AI era, Ismael called on the industry to embed ZK into the heart of machine intelligence, arguing that ZK is not just another crypto tool, but the foundation for the next wave of AI innovation.
“The intersection of ZK / AI is one of the fastest growing areas in crypto today. Lagrange pioneered the ZK / AI space with DeepProve, the fastest zkML library available in the industry.
DeepProve is already being used by dozens of teams ranging from small crypto projects to major FAANG companies for cases. Through our partners, DeepProve will establish a new frontier of ZK use cases including applications from verifying AI agents within DeFi protocols to combatting deepfakes at the nation-state level.
In Dubai, we cohosted the first ZK / AI Summit with House of ZK, bringing together industry leaders in both fields for the first time. We were overwhelmed by the outpouring of support across the over 3000 registrants and 100,000 viewers who tuned into the livestream.”
- Ismael Hishon-Rezaizadeh, Founder & CEO of Lagrange
The summit highlighted verifiability as the missing pillar of today’s AI infrastructure. In a world increasingly shaped by opaque models and unverifiable outputs, zero-knowledge proofs were championed as the key to restoring user trust - not just through open-source models, but through cryptographic guarantees of correctness, provenance, and fairness. One panel during the summit particularly stressed the need to go beyond explainability, framing ZK as a tool for mathematical trust, especially in high-stakes environments like medicine, finance, and digital identity.
Importantly, the summit didn’t just dwell in theory, it grappled with real-world constraints and unanswered questions. What happens when the assumptions behind ZKML break under production-scale workloads? What infrastructure is still missing? How can developers avoid fragmentation while maintaining competitive diversity? Speakers called for stronger abstraction layers, benchmarkable tooling, and developer-centric UX to make verifiable AI not just powerful, but usable. Several projects also advocated for generalized ZK infrastructure that works across all blockchains, not just Ethereum, in order to support decentralized, agentic AI apps with complex, on-chain identity and computation needs.
Privacy was redefined as a fundamental design principle, not just an afterthought. Projects like Self Protocol demonstrated how modular, ZK-enabled data disclosures allow for selective identity verification and anonymous civic participation, without exposing unnecessary personal data. Innovations in federated learning, including peer-to-peer gradient merging without central servers, pushed forward a vision of AI that’s personal, private, and locally trained. From NFC-chip passports to encrypted audits in healthcare, privacy-preserving architectures are no longer a nice-to-have - they’re becoming prerequisites.
Despite the technical rigor, the summit also embraced a broader philosophical and economic perspective. Key questions emerged: Who owns the 'intelligence layer' in the new AI stack? How can decentralized systems prevent capture by centralized actors? Several panelists urged investors and founders to resist narrative-driven hype and instead focus on modular, cryptographic systems that empower user sovereignty. AI models were described not only as tools, but as economic primitive, programmable assets that could be tokenized, composable, and integrated into emerging derivative markets. As infrastructure continues to develop, ZK is poised to secure not just the logic of AI systems, but their value flows and incentives.
While over 20 teams are now tackling overlapping ZKML problems, summit speakers embraced this “overflow” effect as a sign of maturity and necessity. Healthy competition is driving innovation into unexpected areas, from proof markets to encrypted data publication protocols. Rather than balkanization, a strong culture of open-source collaboration remains, a testament to the shared belief that the only sustainable future is one that is verifiable, inclusive, and permissionless by default.
As the summit concluded, one message rang clear: ZK and AI are no longer parallel conversations, but converging forces shaping a shared frontier. Building provable, privacy-preserving, decentralized intelligence is no longer a theoretical aspiration, but an active, accelerating movement now.
The ZK/AI Summit is a global summit initiated by House of ZK, dedicated to exploring the intersection of zero-knowledge and artificial intelligence. By convening leaders and builders across the industry, the summit fosters open collaboration toward a verifiable, privacy-preserving future of AI with ZK.
House of ZK is a platform dedicated to exploring and promoting the latest advancements in the Zero-Knowledge industry, designed to be a one-stop hub for anyone interested in ZK, and consists of three main pillars: Podcast, Journal, and Events. Each pillar is uniquely curated to provide exclusive insights about the cutting-edge developments in the industry, ultimately providing a communal space where ideas are shared, discussed, and developed.
📧 For collaborations, please contact: contact@hozk.io
Lagrange Labs builds DeepProve, is the fastest and most scalable verifiable Inference library for AI Models.
As AI systems become increasingly embedded in our daily lives, ensuring their trustworthiness and transparency is paramount. In critical sectors such as finance, healthcare, and defense, the integrity of AI-driven decisions is not just desirable but essential. The opaque nature of many AI models raises concerns about the correctness of their outputs and the potential for tampering or misuse.
DeepProve addresses these challenges by enabling verifiable AI inference. Our library allows stakeholders to cryptographically confirm that a neural network was executed as intended, ensuring both the correctness of the computation and the integrity of the input data, all without exposing sensitive information. Unlike existing solutions, DeepProve achieves proof generation speeds that 1000x faster, making it suitable for real-world applications where performance and privacy are critical.
DeepProve is integrating with top teams across AI and DeFi including Sentient, Mira, ZKsync, Polygon, Gaia, OpenLedger, LayerZero, Bio Protocol, 3Jane and Frax.
