In this exclusive Cryptonomist interview, DFINITY explores how the Internet Computer (ICP) is building the World Computer, ushering in a new era of scalable, onchain AI, data sovereignty, and Web3 interoperability.
1. Can you explain the concept of the “World Computer” on ICP and how it differs from other blockchain-based platforms in terms of accessibility, scalability, and sovereignty?
The Internet Computer hosts secure compute at scale, but with the World Computer, we aim to further bind together different decentralized networks, creating unprecedented accessibility, scalability, and data sovereignty. In terms of accessibility, apps are hosted fully onchain and accessible through standard URLs, creating a seamless user experience, similar to familiar web2 apps. Canister smart contracts are not as limiting as traditional smart contracts, enabling new levels of scalability – storing large datasets and rapidly processing requests. Due to all nodes running independently, and with no centralized party controlling the network, the World Computer built on ICP enables total user data sovereignty.
2. What are the main challenges in building a fully onchain system that can match the scalability and user experience of Web2 applications? How is ICP addressing these challenges?
The typical challenges of building an onchain network that we have noticed include scalability, user experience, and data and compute limitations.
ICP addresses these challenges by using canister smart contracts, enabling larger data storage than traditional smart contracts; also storing everything onchain with no external cloud server hosting the data. ICP does not build on web2 infrastructure like AWS, we are building the internet stack itself enabling the web2 experience to perform entirely onchain – ensuring a comfortable user experience, similar to web2 but with enhanced security.
3. The integration of AI models on smart contracts is a unique approach. How does DFINITY ensure the security and tamperproof nature of these AI models while maintaining scalability?
ICP’s AI model deployment takes place entirely onchain, and the model’s logic is immutable unless designed to be upgradeable. That means that all data inputs and outputs are transparent, ensuring that the model cannot be tampered with without detection. This aids in preventing any malicious behaviour from taking place.
Each canister exists and runs in a sandbox environment, thus they are isolated from other parts of the system, meaning that any incidents will be contained. Orthogonal persistence also allows canisters to maintain their state no matter the conditions, allowing AI services to run without needing to reload or reinitialize after every interaction.
4. Can you share some practical use cases where ICP’s onchain AI models could significantly impact industries or user experiences?
ICP’s onchain AI models enable highly secure and transparent decentralized applications that have the potential to transform user experiences across industries – from healthcare, to financial services and education. AI can offer fraud detection, verify authenticity and traceability in supply chains, and even create personalised tutors. Due to their nature as smart contracts, AI agents built on ICP do not rely on intermediary centralized systems, ensuring tamperproof and verifiable outputs – boosting reliability, and ensuring open access and data sovereignty.
In terms of a product using AI onchain that we are very excited about, ICP will soon launch a platform called Caffeine. This is our proprietary AI app building tool that will enable a complete paradigm shift towards the ‘self-writing internet’. Using natural-language prompts in a seamless user interface, Caffeine will allow anyone, anywhere, to build, deploy and update any app they desire, without the need for complex coding knowledge or expensive technical equipment. This is truly the next step in making AI onchain both tangible and accessible for all.
5. What specific aspects of Web3 applications can benefit the most from combining AI with blockchain technology, and how does DFINITY envision this integration evolving?
Autonomous agents and smart services – ICP-enabled AI can extend the logic of smart contracts, enabling intelligent agents that can learn in real-time, reason and also adapt. These agents can manage DeFi portfolios or automate DAO governance processes based on real-time data and behaviour. On ICP specifically, these can be executed within canister smart contracts, which are secure and auditable directly on chain. In addition, AI models can support personalization in decentralized apps by analyzing preferences and behaviour to deliver tailored content. This has previously been specific to Web2 and lacking in Web3, but that will now change.
6. As a former IBM Fellow with extensive experience in secure multiparty computation, how has your background influenced DFINITY’s approach to privacy and security in onchain AI models? (For Jan)
My background in cryptographic protocols, particularly secure multiparty computation and privacy-enhancing technologies, has significantly shaped DFINITY’s architectural approach to both privacy and security, especially as we work to explore the integration of AI models into the blockchain layer.
One of the core principles that guided the development of ICP was that privacy and decentralization should not be in tension but instead must co-exist. This belief stems directly from my work in designing cryptographic protocols that enable collaborative computation without sacrificing data confidentiality.
We leverage chain-key cryptography, a novel technique which enables secure, single-public-key authentication. This allows smart contracts to serve web content, manage identity and run AI without reliance on external providers. Our canister architecture also enforces isolation and access at the computation layer. AI models can be embedded within the canisters, and their interaction with user data is fully governed by on-chain logic, which can be inspected and audited. We encourage users to control how their identity and data are managed across applications, thus we developed Internet Identity to provide strong, cryptographically protected authentication.
7. With your experience in strategy consulting and venture building, how do you see DFINITY positioning itself in the evolving landscape of decentralized technology? (For Pierre)
We focus on the core USPs that ICP brings to the table:
Digital Sovereignty: This is no longer just a niche concern—it’s a matter of national resilience. Over 90% of cloud compute is controlled by providers governed by foreign intelligence laws, making “tech colonialism” a rising issue. At the same time, the global spend on cybersecurity exceeds USD 9 trillion. ICP offers the solution: a decentralized full-stack compute infrastructure that goes far beyond the definition of web3 as a token ledger – web3 means full stack sovereign compute (frontend, backend, data replicated and secured by cryptography). It enables the creation of sovereign clouds and apps, completely free from foreign interference and immune to cyberattacks. This is the only viable path to true digital sovereignty in today’s world.
Interoperability: We recognize two key points for widespread web3 adoption: (a) the complexity of switching between blockchain networks must be simplified for users, and (b) developers, enterprises, and governments must have the flexibility to interact with multiple blockchain ecosystems without being locked into one. ICP’s Chain Fusion technology addresses both. Built from the ground up to support interoperability, ICP enables seamless interaction with networks like Bitcoin, Ethereum, and Solana—and many more in the future. This allows users to interact with any relevant blockchain ecosystem but also vice versa allows builders in other blockchain ecosystems to benefit from ICP’s decentralized compute, extending the power of full-stack decentralization beyond token-ledgers.
AI: ICP is not only capable of hosting large language models (LLMs) and AI agents on a decentralized infrastructure; it also serves as the AI cloud for self-writing apps. AI on ICP empowers anyone to build, deploy, update, and orchestrate their own applications through natural language prompting. This is a new paradigm for the entire software industry and a new use case where a true web3 infrastructure is superior to legacy IT. This is only possible on the Internet Computer and will reshape the dynamics and rules of the software industry.
All of these use cases will feel like Web2 interactions to users—seamless, intuitive, and without the friction currently seen in Web3.
8. Could you elaborate on the importance of data sovereignty in the context of decentralized AI models and how ICP aims to address user concerns regarding data ownership and control?
Data sovereignty is crucial in decentralized AI systems for integrity, transparency and trustworthiness – there are currently a lot of underlying concerns about the privacy and potential misuse of data, so getting this right is mission critical.
Decentralized AI models, especially those built on ICP’s platform, are engineered to put data control back in the hands of the users and also ensure complete transparency. For instance, canisters can enforce sophisticated control policies that specify how, when and by whom data is accessed. These rules are immutable and auditable, giving users and enterprises confidence that their data is completely protected. This has become increasingly important with cybercrime on the rise – with attacks becoming more frequent and sophisticated.
9. What upcoming developments can we expect from DFINITY that will further enhance the capabilities of the World Computer on ICP?
DFINITY has many exciting developments in the pipeline for the coming year and beyond. We recently announced our ICP Ninja upgrade, transforming developer experiences with a focus on collaboration and productivity. Our DeAI ecosystem continues to grow, with the recent launch of the Alpha version of Caffeine, allowing prompt-based creation of full-stack applications. We have many exciting partnerships coming alongside our extensive roadmap developments – watch this space!
10. Looking ahead, how do you envision the role of decentralized AI evolving in the context of broader Web3 adoption, and what role will ICP play in that future?
As data sovereignty, trustless computation and censorship resistance all become more pertinent, decentralized systems will provide an excellent alternative to the traditional centralized models that rely on outdated infrastructure that are governed by third parties. Web3 will host applications that will natively integrate AI agents that can interact and adapt autonomously, all while operating under transparent and self-owned governance.
ICP is well placed to support this evolution, as unlike other Layer-1 blockchain ecosystems that rely on off-chain infrastructure, ICP operates directly onchain. Through ICP’s smart contracts, developers can build AI services that are fully decentralized, tamper-proof, and interoperable with various other Web3 services.
