a16z: 11 major use case insights on the intersection of AI and encryption

Source: a16z crypto; Translation: Golden Finance xiaozou

The economic model of the internet is undergoing a transformation. When the open internet collapses into a simple prompt box, we can't help but ask: Will AI lead us to an open internet, or will it construct a new maze of paywalls? And who will control all of this, highly centralized giant corporations, or the vast user community?

This is precisely the field where cryptographic technology shines. We have discussed the intersection of AI and cryptographic technology multiple times. In short, blockchain provides a new paradigm for building internet services and networks—decentralized, trust-neutral, and truly user-owned. By restructuring the economic model of current systems, blockchain can effectively counter the increasingly evident centralization trend in AI systems, helping to create a more open and reliable internet.

Cryptography can optimize AI systems, and vice versa—this idea is not new, but it is often vaguely defined. Certain cross-domain areas have attracted builders and users, such as validating "human proof" in the current era of low-cost AI proliferation. However, other use cases still seem to require years or even decades to realize. In this article, we will share 11 practical use cases of the fusion of cryptography and AI, aiming to stimulate discussions about feasibility, challenges to be solved, and more. These cases are all based on technologies currently under development, ranging from processing massive micropayments to ensuring human control over the relationship with future AI.

1**,AI persistent data and context in interaction**

Author: Scott Duke Kominers, a16z crypto research partner

Generative AI relies on data to thrive, but for many applications, the importance of context (i.e., the state and background information related to interactions) is no less than that of data, and may even be more critical.

Ideally, AI systems (whether they are agents, large language model interfaces, or other applications) should be able to remember a wealth of details such as the types of projects you are working on, your communication style, and your preferred programming languages. However, in reality, users often need to re-establish this context in different interactions within a single application (for example, each time they start a new ChatGPT or Claude session), not to mention when switching across systems.

Currently, the context of generative AI applications is almost impossible to transfer between different systems.

By leveraging blockchain technology, AI systems can transform key contextual elements into persistent digital assets. These assets can be loaded at the start of a conversation and seamlessly transmitted across AI platforms. More importantly, blockchain may be the only solution that simultaneously offers commitments to both forward compatibility and interoperability—because these features are the core attributes based on blockchain protocols.

The gaming and media sectors are natural application scenarios: user preference settings can persist across different games and environments. However, the real value lies in the field of knowledge application (AI needs to understand the user's knowledge structure and learning style) and specialized AI use cases like programming. Of course, companies have already begun developing customized robots with business-specific global contexts—but such contexts are often not transferable, and even different AI systems used within the same organization cannot share them.

Institutions have just begun to realize this issue. The closest universal solution at present is customized robots with fixed persistent contexts. However, the portability of context among users within the platform has begun to sprout off-chain: for example, the Poe platform allows users to rent out their customized robots.

Bringing such activities on-chain will enable our interactive AI systems to share a contextual layer that includes all key elements of digital activities. They will be able to immediately understand our preferences and optimize the user experience more accurately. Conversely, just as the on-chain intellectual property registration mechanism allows AI to reference persistent on-chain context, it can also give rise to new market interactions around prompts and information modules — for example, users can directly authorize or monetize their expertise while retaining control over their data. Of course, sharing context will also enable many possibilities that we have yet to envision.

2**, Universal Identity System for Intelligent Agents**

Author: Sam Broner, Partner at a16z crypto investment team

Identity—an authoritative credential that records the essence of things—is the invisible infrastructure supporting today's digital discovery, aggregation, and payment systems. Since platforms enclose this infrastructure within walls, the identity we perceive is only part of the finished product: Amazon assigns identifiers (ASIN or FNSKU) to products, centrally displays them, and assists users in discovery and payment. The same goes for Facebook: user identity forms the basis of its information flow, permeating all scenarios within the app, including Marketplace product listings, organic posts, and paid advertisements.

With the development of AI agents, all of this is about to change. As more companies use agents for customer service, logistics, payments, and other scenarios, their platforms will increasingly resemble not a single interface application but will operate across multiple carriers and platforms, accumulating deeper context to perform more tasks for users. However, if the identity of the agent is bound to a single market, it will be unusable in other important scenarios (email threads, Slack channels, within other products).

Therefore, agents need a unified "digital passport". Without it, we will not be able to pay agent fees, verify their versions, query their functions, confirm their agents, or track their cross-application reputation. Agent identity needs to encompass wallet, API registry, update logs, and social credentials functions—enabling any interface (email, Slack, or other agents) to recognize and interact with them in a unified manner. The lack of a shared foundational element of "identity" means that each integration needs to rebuild the infrastructure from scratch, and the discovery mechanisms are always in a temporary state, losing context when users switch channels.

We have the opportunity to design agent infrastructure from first principles. So how do we build a more robust and trustworthy neutral identity layer than DNS records? Rather than recreating an integrated platform that bundles identity with discovery, aggregation, and payment, we should allow agents to receive payments, specify functions, and exist across multiple ecosystems without being constrained by a specific platform. This is precisely where the value of the intersection of cryptography and AI lies—blockchain networks provide permissionless composability, enabling developers to create more useful agents and a better user experience.

Currently, vertically integrated solutions such as Facebook or Amazon indeed provide a better user experience—the inherent complexity of creating excellent products includes ensuring coordination and unity from top to bottom across all links. However, the cost of this convenience is high, especially as the costs of building agent aggregation, marketing, monetization, and distribution software decrease, while the application scenarios for agents continue to expand. Although matching the user experience provided by vertically integrated suppliers still requires effort, a trustworthy and neutral agent identity layer will truly allow entrepreneurs to own their digital passports and encourage them to innovate boldly in the fields of distribution and design.

3**, forward-compatible proof-of-humanity mechanism**

Author: Jay Drain Jr., a16z crypto investment partner; Scott Duke Kominers, a16z crypto research partner

As AI increasingly permeates various online interactions (from deep fakes to social media manipulation, supporting various bots and agents), it becomes increasingly difficult to discern whether online interaction partners are real humans. This trust crisis is not a future risk but a present reality—blurring the lines between the real and the virtual, from the army of comments on the X platform to bots on dating apps. In this environment, human verification becomes a key infrastructure.

Digital IDs (including centralized IDs used by the U.S. Transportation Security Administration) are a way to verify human identity. These IDs contain all credentials that can prove one's identity, such as usernames, PIN codes, passwords, and third-party authentication (like citizenship or credit ratings). The value of decentralization is evident here: when this data is stored in a centralized system, the issuer can revoke access, charge fees, or assist in monitoring at any time. Decentralization completely reverses this power structure: users, rather than platform gatekeepers, control their own identities, making them more secure and resistant to censorship.

Unlike traditional identity systems, decentralized human proof mechanisms (such as Worldcoin's Proof of Human) allow users to self-custody and verify their human identity while protecting privacy and maintaining trust neutrality. Just as a driver's license is available everywhere regardless of the time and place of issuance, decentralized PoP (Proof of Personhood) can serve as a reusable foundational layer for any platform, including platforms that have yet to emerge. In other words, blockchain-based PoP has forward compatibility because it possesses:

Portability: The protocol is available as a public standard for integration on any platform. Decentralized PoP is managed through public infrastructure and is fully controlled by users. This provides complete portability, allowing compatibility with any platform now or in the future.

Permissionless Accessibility: The platform can autonomously choose to acknowledge PoP ID without going through gatekeeping APIs that may discriminate against different use cases.

The challenge faced in this field is the adoption rate. Although large-scale human proof application cases have not yet emerged, we expect that the critical user mass, early partners, and killer applications will accelerate its popularity. Each application that adopts a specific digital ID standard will enhance the value of that ID to users, thereby attracting more users to acquire that ID, which in turn will encourage more applications to integrate that ID as a means of human authentication (since on-chain IDs inherently possess interoperability, this network effect can form rapidly).

We have seen mainstream consumer applications in the fields of gaming, socializing, and social media announce partnerships with World ID to help users verify their identities when playing, chatting, and trading with real humans (and specific intended parties). This year has also seen the emergence of new identity protocols such as the Solana Attestation Service (SAS) — although it does not directly issue human proofs, SAS allows users to associate off-chain data (such as KYC checks or investment qualification certificates required for compliance) with Solana wallet privacy, thus building decentralized identities. These signs indicate that the turning point for decentralized PoP may not be far off.

Human proof is not only about prohibiting robots but also about defining a clear boundary between AI agents and human networks. It enables users and applications to distinguish between human-computer interactions, creating space for a higher quality, safer, and more authentic digital experience.

4**, Decentralized Physical Infrastructure Network oriented towards AI ( DePIN )**

Author: Guy Wuollet, Partner at a16z crypto investment team

Although AI belongs to digital services, its development is increasingly constrained by the bottleneck of physical infrastructure. The decentralized physical infrastructure network ( DePIN ) - a new model for building and operating physical systems - can help popularize the computing infrastructure necessary for AI innovation, making it more cost-effective, resilient, and better at resisting censorship.

How to achieve this? The acquisition of energy and chips are the two core obstacles to the development of AI. Decentralized energy can enhance power supply, while builders are also integrating idle chips from scenarios such as gaming PCs and data centers through DePIN. These computers can collectively form a permissionless computing resource market, creating a fair competitive environment for the development of new AI products.

Other application scenarios include distributed training and fine-tuning of large language models, as well as distributed networks for model inference. Decentralized training and inference can significantly reduce costs as it utilizes previously idle computing resources. At the same time, it provides censorship resistance, ensuring that developers are not deprived of platform access by extremely large cloud service providers (centralized cloud service giants that offer elastic computing infrastructure).

The issue of AI models being concentrated in a few companies has long existed; decentralized networks help create more cost-effective, censorship-resistant, and scalable AI systems.

5**,AI intelligent agents, the infrastructure and protection mechanisms for interaction between terminal service providers and users**

Author: Scott Duke Kominers, a16z crypto research partner

As the capabilities of AI tools to perform complex tasks and multi-level interaction chains improve, the demand for autonomous interaction among agents will significantly increase.

For example, an AI agent may need to acquire specific computational data, or call on a specialized agent to perform a specific task—such as assigning a statistical bot to develop and run a model simulation, or enabling an image-generating bot in the production of marketing materials. AI agents can also create tremendous value by completing the entire transaction process on behalf of the user – such as searching for and booking flights based on preferences, or discovering and ordering a new type of book.

Currently, a universal inter-agent market has not yet been formed, and such cross-system queries are mainly achieved through explicit API connections or are limited to closed ecosystems that support internal agent calls.

Broadly speaking, most AI agents currently operate in isolated ecosystems, with APIs being relatively closed and lacking architectural standardization. Blockchain technology can help establish open standards for protocols, which is crucial for short-term adoption. In the long run, it also supports forward compatibility: when new AI agents emerge, they can seamlessly integrate into existing underlying networks. Thanks to its characteristics of interoperability, open-source, decentralization, and easy upgrades, blockchain can better adapt to AI innovation and iteration.

With the development of the market, several companies have begun to build blockchain infrastructure for interaction between intelligent agents: for example, Halliday recently launched a standardized cross-chain architecture protocol that supports AI workflow interaction, ensuring that AI behavior does not deviate from user intent through protocol-level protection. Catena, Skyfire, and Nevermind utilize blockchain to achieve autonomous payments between agents without the need for human intervention. More similar systems are being developed, and Coinbase has even begun to provide infrastructure support for these attempts.

6**, maintain the synchronization of AI/atmosphere coding applications**

Author: Sam Broner, Partner at a16z crypto investment team; Scott Duke Kominers, Research Partner at a16z crypto

The revolutionary advancements in generative AI have made software development unprecedentedly simple. Coding efficiency has improved by orders of magnitude, and more importantly—programming can now be done in natural language, allowing even inexperienced developers to fork existing programs or build new applications from scratch.

However, while AI-assisted coding creates new opportunities, it also introduces a significant amount of entropy both within and outside the program. "Vibe coding" abstracts away the complex dependency networks at the software's core, but as source libraries and other inputs change, this programming approach may lead to functional and security vulnerabilities in the program. Additionally, when people use AI to create personalized applications and workflows, the difficulty of integrating these systems with others also increases. In fact, two vibe coding programs executing the same task may have completely different operational logic and output structures.

For a long time, the standardization work to ensure consistency and compatibility was initially undertaken by file formats and operating systems, and later realized through shared software and API integration. However, in a world where software evolves, morphs, and forks in real time, the standardization layer needs to have broad accessibility and continuous upgradability while maintaining user trust. More importantly, relying solely on AI cannot solve the problem of incentivizing people to establish and maintain these links.

Blockchain technology can simultaneously solve these two problems: embedding user-customized software construction through protocolized synchrony layers, dynamically updating to ensure cross-platform compatibility in a changing environment. In the past, large enterprises might spend millions of dollars hiring "system integrators" like Deloitte to customize Salesforce instances, but today engineers can create customized interfaces to view sales information in just one weekend. However, with the explosion of customized software, developers need assistance to keep these applications running in sync.

This is similar to the current development model of open source software libraries, but the difference lies in continuous updates rather than periodic releases—along with an added layer of incentives. These two points are more easily realized under the support of cryptographic technology. As with other blockchain-based protocols, the co-ownership of the synchronization layer motivates all parties to continuously invest in improvements. Developers, users (and their AI agents), and other participants can be rewarded for introducing, using, and developing new features and integrations.

Conversely, shared ownership aligns all users with the overall success of the protocol, serving as a buffer mechanism against malicious behavior. Just as Microsoft would not undermine the .docx file standard to avoid impacting users and brand reputation, the co-owners of the synchronization layer similarly have no incentive to introduce poor or malicious code into the protocol.

Like all previous standardized software architectures, there is enormous potential for network effects in this field. As the "Cambrian explosion" of AI coding software continues, the need to maintain interconnected heterogeneous system networks will expand dramatically. In short: ambient coding must remain in sync and cannot rely solely on Vibe. Cryptography is precisely where the answer lies.

7**, a micropayment system that supports revenue sharing**

Author: Liz Harkavy, Partner at a16z crypto investment team

AI tools and agents represented by ChatGPT, Claude, and Copilot provide a new convenient way to explore the digital world. However, regardless of the pros and cons, they are shaking the economic foundation of the open internet. The real impacts are already evident—educational platforms are facing a sharp decline in traffic due to students turning to AI tools, and several U.S. newspapers are suing OpenAI for copyright infringement. If we cannot restructure the incentive mechanisms, we will witness an increasingly closed internet, with more paywalls and fewer content creators.

While policy measures certainly exist, multiple technical solutions are emerging alongside the advancement of legal procedures. The most promising (and also the most complex) solution may be to embed a revenue-sharing system within network architecture: when AI-driven actions facilitate transactions, the content sources involved in the decision-making process should receive a share. Affiliate marketing ecosystems have already achieved similar attribution tracking and revenue distribution, and more advanced versions could automatically track all contributors in the information chain and provide rewards—blockchain can clearly play a role in the traceability chain.

However, such systems require new infrastructure equipped with special functions: micro-payment systems that can handle multi-source microtransactions, attribution protocols that fairly assess various contributions, and governance models that ensure transparency and fairness. Existing blockchain tools have shown potential, such as Rollup and Layer2 solutions, AI-native financial institution Catena Labs, and financial infrastructure protocol 0xSplits, which can achieve near-zero cost transactions and more refined payment splits.

Blockchain will empower smart payment systems through the following mechanisms:

• Nano payments can be split among multiple data providers, allowing for automatic distribution of very small payments to all contributing sources through smart contracts with a single user interaction.

• Smart contracts support executable traceable payments based on completed transactions, compensating information sources that influence purchasing decisions after transactions occur in a fully transparent and traceable manner.

• Supports complex programmable payment allocation schemes, achieving fair profit distribution through code-enforced rules rather than centralized decision-making, establishing trustless financial relationships between autonomous agents.

As these emerging technologies mature, they will create new economic models for the media industry that capture the entire value chain—from creators to platforms to users.

8**, as a blockchain for intellectual property and traceability registry**

Author: Scott Duke Kominers, a16z crypto research partner

The rise of generative AI urgently calls for efficient and programmable mechanisms for intellectual property registration and tracking—ensuring traceability of content sources while supporting business models around IP access, sharing, and remixing. The current IP framework relies on costly intermediaries and post-facto accountability, which can no longer adapt to the new era of AI instant content consumption and one-click variant generation.

We need an open and transparent registration system that can provide clear proof of ownership, allowing IP creators to interact conveniently and efficiently, and enabling AI and other web applications to connect directly. Blockchain is a perfect solution: it allows IP registration to be completed without intermediaries, provides immutable provenance proof, and enables third-party applications to easily identify, authorize, and call these IPs.

As the first two eras of the internet (and the ongoing AI revolution) are often associated with the weakening of intellectual property protection, the idea that technology can protect IP naturally raises many doubts. The issue is that most current IP business models focus on excluding derivative works rather than incentivizing and monetizing these creations. However, programmable IP infrastructure not only allows creators, franchisees, and brands to clearly define IP ownership in the digital space but also opens the door to IP-sharing business models centered around digital applications like generative AI—transforming the main threat of generative AI to creative work into an opportunity.

We have seen creators experimenting with new models in the early stages of the NFT space, with companies leveraging NFT assets on Ethereum to achieve network effects and value accumulation under CC0 brand building. Recently, there have even been protocols specifically designed for the registration and licensing of standardized, composable IP, as well as dedicated blockchains (such as Story Protocol). Some artists have begun to authorize their artistic styles and works for creative remixing through protocols like Alias, Neura, and Titles. Meanwhile, Incention's Emergence series allows fans to participate in the co-creation of sci-fi universes and characters, tracking the contributions of each creator through a blockchain registry built on Story.

9**, web crawlers that help content creators monetize**

Author: Carra Wu, Partner at a16z crypto investment team

The AI agent that currently has the best product-market fit is not a programming or entertainment assistant, but a web crawler - these digital agents autonomously traverse the internet, collect data, and determine link tracking paths.

It is estimated that nearly half of all web traffic already originates from non-human subjects. Bots routinely ignore robots.txt protocols (the app is weak enough to inform automated crawlers of access), and the data they collect ends up being a competitive barrier for some tech giants. To make matters worse, websites have to bear the cost of bandwidth and CPU resources for these uninvited guests, as if serving a never-ending anonymous data harvester. The blocking solutions provided by CDN (Content Delivery Network) service providers such as Cloudflare are actually remedies that should not exist in the first place.

We've pointed out that the original contract of the Internet – the economic pact between content creators and distribution platforms – is about to collapse. The data confirms this trend: in the past 12 months, website owners have started banning AI crawlers on a large scale. In July 2024, only about 9% of the world's top 10,000 websites blocked AI crawlers, and now the proportion has reached 37%. This number is bound to continue to climb as the main defense methods of the website are upgraded and user dissatisfaction accumulates.

If we do not rely on CDN to completely block suspected crawler visitors, can we find a compromise solution? Instead of abusing systems designed for human traffic, AI crawlers could potentially pay for data collection rights. This is where blockchain comes into play: in this scenario, each crawler agent would hold cryptocurrency and negotiate on-chain with the website's "bouncer" agent or paywall protocol through the x402 protocol (of course, the challenge lies in the fact that the Robots Exclusion standard, which has been in use since the 1990s, is deeply rooted and would require large-scale collective cooperation from CDN giants like Cloudflare to overcome).

At the same time, human users can verify their real identity through World ID (refer to the previous text) to continue accessing content for free. In this way, content creators and website owners can receive reasonable compensation for AI training sets at the data collection stage, while humans can still enjoy a free-flowing information internet.

10**, A New Paradigm of Advertising that is Both Precise and Private**

Author: Matt Gleason, a16z crypto security engineer

AI has begun to change our online shopping experience, but what if the ads we see daily could actually be useful? The reasons people dislike ads are obvious: irrelevant ads are nothing but noise, while overly targeted AI ads (based on massive consumer data) can be creepy. Other applications monetize through unskippable ad walls (such as streaming services or game levels).

Cryptographic technology can reconstruct advertising mechanisms and address these pain points. By combining blockchain with personalized AI agents, a balance can be found between "irrelevant ads" and "terrifying precision"—ads can be delivered based on user-defined preferences. The key is that all of this can be done without globally exposing user data, and it can directly compensate data sharers or ad interactors.

The required technical elements include:

• Low-Fee Digital Payments: To compensate users for their ad interactions (views/clicks/conversions), businesses need to frequently send small payments. This requires a high-throughput, near-zero cost payment system.

• Privacy-Protected Data Verification: AI needs to demonstrate that consumers meet certain demographic characteristics. Zero-knowledge proofs can complete the verification while protecting privacy.

• Incentive Mechanism: If the internet adopts a micropayment-based monetization model (such as interactions costing less than $0.05), users can choose to watch ads in exchange for rewards, transforming the current "extraction model" into a "participation model".

Humans have been pursuing advertising relevance for hundreds of years (offline) and decades (online). By reconstructing advertising through the lens of crypto and AI, it will ultimately become truly useful: precise yet not intrusive, achieving a win-win situation for all parties—unlocking a more sustainable and aligned incentive structure for builders and advertisers; and providing users with more ways to explore the digital world.

This will not only not depreciate the value of advertising space, but will instead enhance its significance. It is also expected to subvert the deeply entrenched exploitative advertising economy, replacing it with a more humane system: users are seen as participants rather than products.

11**, AI**** partners owned and controlled by humans**

Author: Guy Wuollet, Partner at a16z crypto investment team

Modern people spend more time on electronic devices than on face-to-face communication, with an increasing amount of time spent interacting with AI models and AI-filtered content. These models essentially provide some form of companionship—whether for entertainment, information acquisition, interest satisfaction, or children's education. It is not hard to imagine that in the near future, AI-based educational assistants, health advisors, legal assistants, and emotional partners will become the mainstream mode of interaction for humans.

Future AI partners will possess infinite patience and be able to deeply adapt to the specific needs of individual users. They are not just assistants or robotic servants; they may evolve into highly valued "relationships." Therefore, the ownership and control of these relationships (whether by users, companies, or other intermediaries) become critically important. If you have been concerned about content moderation and censorship on social media over the past decade, this issue will become exponentially more complex and personal in the future.

Censorship-resistant hosting platforms (such as blockchain) provide the strongest path for enabling user-controlled, uncensorable AI—this is not a new argument (as previously discussed). While individuals can run local models or purchase GPUs, most either cannot afford it or lack the technical ability.

Although the widespread adoption of AI companions still requires time, the relevant technology is rapidly evolving: text-based humanoid companions have become quite mature, virtual avatars have significantly improved, and blockchain performance continues to enhance. To ensure the usability of censorship-resistant assistants, a better user experience in crypto applications is essential. Fortunately, wallets like Phantom have greatly simplified blockchain interactions, and embedded wallets, passkeys, and account abstraction technology allow users to self-custody wallets without having to remember seed phrases. With high-throughput trustless computers (using technologies like optimistic proofs and ZK co-processors), establishing meaningful and lasting relationships with digital companions will become possible.

In the near future, the focus of discussion will shift from "When can we see realistic digital companions" to "Who can control them and how to control them."