Privacy as a primitive
Crypto proved it can make finance transparent. But can it make it private? As AI agents become financial actors, public-by-default rails turn into attack surfaces. This issue explains why privacy must become a base-layer primitive.
Crypto markets are transparent by design. That transparency has been treated as a virtue — auditability, verifiability, trust minimization. But transparency and privacy are often in tension, and for most of crypto’s life, privacy has not been a base-layer assumption. It has been an afterthought.
That design choice made sense when crypto was primarily a market. Markets like public tapes. Price discovery thrives on visibility, and so does trustless verification.
But if crypto is going to be money — payroll, invoices, treasury operations, autonomous agents negotiating and settling value — then privacy stops being philosophical and starts being structural. This is true at the individual payments level. It is even more true at the institutional level.
Daniel McGlynn
This week’s framing is simple: privacy isn’t a feature. It’s a missing primitive. And the pressure to make it native is coming not from ideology, but from scale — especially as AI agents begin to act as financial participants. Just like you wouldn't want a Venmo payment to expose your bank account balance, you wouldn't want a basic everyday crypto transaction to reveal other financial details or behaviors.
What it means for privacy to be a primitive
A primitive is something the rest of the stack can assume exists. Like TCP/IP are the base rails for the internet. Or like how programmable transfers work on Ethereum.
In crypto today, privacy is not a primitive. It’s a workaround. If you want confidentiality, you typically:
- exit to a specialized tool
- rely on obfuscation rather than cryptographic guarantees
- trust a centralized intermediary
- or hope your activity is too small to attract scrutiny
All of those approaches create friction for an individual user and none of those approaches scale to enterprise finance, institutional DeFi, or machine-driven capital flows.
The regulatory history reinforces the tension. The U.S. Treasury’s sanctions against Tornado Cash — and later reversal — showed how unsettled the legal treatment of privacy infrastructure remains (Reuters, March 2025: U.S. scraps sanctions on Tornado Cash). The narrative around privacy seems to be that private transactions are only useful tools for criminals.
What that episode demonstrated is that privacy tools operating without clearly legible compliance hooks trigger political backlash. The state is still working out how to handle infrastructure that obscures financial flows, especially when that infrastructure is open-source and non-custodial.
Daniel McGlynn
Why crypto hasn’t had privacy
There’s a common explanation: builders prioritized scalability and composability first.
That’s true, but the deeper reasons are structural. Saying privacy was sacrificed for speed of building or adoption isn't really the full story.
Transparency became the compliance default. Public ledgers make surveillance cheap, which benefits exchanges, regulators, and analytics firms. Chain analysis became a growth industry precisely because blockchains are legible.
Composability prefers determinism. DeFi works because contracts can inspect state openly. If contracts cannot easily see balances and flows, you lose some of the Lego-like interoperability that fueled early DeFi growth.
Markets reward visibility. Most crypto activity has centered around trading. Trading benefits from observable liquidity and order flow.
The system optimized around what it already was: a public market.
That optimization had consequences. Crypto normalized the idea that radical transparency was synonymous with trust. In reality, it was simply the easiest way to reconcile decentralization with compliance.
What’s changing now
Two shifts are forcing privacy back into the conversation.
First, institutional adoption is colliding with public transparency. Enterprise stablecoin guides increasingly flag confidentiality as a blocker for payroll and treasury use cases (Alchemy enterprise guide: Enterprise stablecoin adoption guide). Real businesses do not want competitors mapping their settlement flows in real time.
Second, AI agents are becoming economic actors, and the transparency that made the early days of crypto possible might make it incompatible with the coming age of agentic commerce.
AI agents change the privacy math
Agentic systems are not theoretical. Financial institutions are openly experimenting with autonomous agents interacting with markets, and regulators are already flagging systemic risks (Reuters: UK regulator warns of risks in agentic AI race).
When agents transact onchain, three technical pressures intensify.
1. Strategy leakage becomes continuous.
On transparent rails, an agent’s transaction graph reveals behavior patterns. MEV and front-running are already endemic in DeFi. When trading logic is automated and persistent, adversaries can model and exploit patterns faster than humans can intervene. The edge shifts toward whoever can parse the public data stream most efficiently.
2. Credential concentration becomes systemic risk.
Agents require programmatic access: private keys, API tokens, contract permissions, scoped authorities. Broader software security research has documented how AI agents frequently request high-privilege credentials in production environments (Diginomica: AI agents requesting high-privilege credentials). In a financial context, those credentials are more than infrastructure access — they are direct claims on capital.
3. Machine-speed exploitation compresses reaction time.
Security research on autonomous agents interacting with smart contracts shows that automated systems can both exploit and be exploited at machine speed (OpenAI: EVM autonomous agent security benchmark). When agents are both counterparties and targets, transparency amplifies the attack surface. Latency becomes destiny.
Transparent markets are challenging. Transparent agentic finance is brittle.
If every move an autonomous treasury agent makes is publicly visible, adversaries do not need to guess. They can simulate, anticipate, and position against it.
Privacy-preserving infrastructure
The technical response is no longer theoretical.
Zero-knowledge proofs (zk) allow participants to prove compliance, solvency, or constraint satisfaction without revealing underlying data. That makes it possible to show that rules are followed without publishing the full transaction graph.
Fully homomorphic encryption (FHE) pushes the idea further: computation on encrypted data, enabling confidential smart contract state while preserving public verification.
The Open Money angle
Open money cannot function as open surveillance.
Real economic life requires:
- confidentiality for payroll and vendor relationships
- selective disclosure for regulators
- protected trading strategies
- bounded authority for automated agents
The frontier is privacy-preserving compliance: systems where participants can prove constraints are met without exposing their entire financial graph.
Stablecoins moving into payroll, institutional DeFi desks running autonomous agents, and AI-native treasury operations all collide with the same constraint: public transparency does not scale to real economic behavior.
Research backlog
- What does privacy with accountability look like technically: zk attestations, FHE-based state, trusted execution environments?
- Which category forces adoption first: enterprise stablecoins, institutional DeFi, or agent wallets?
- How does MEV evolve in a world of encrypted transaction flows?
- What regulatory frameworks tolerate selective disclosure primitives?
- Does privacy live at the base layer, or emerge at the wallet and application layer?
Closing thought
Crypto spent a decade proving it could make finance transparent.
The next decade may hinge on whether it can make finance selectively opaque.
If AI agents are going to hold keys, route capital, and negotiate contracts on our behalf, privacy will be operational infrastructure.
Open money can be programmable and permissionless. But without native privacy, it risks becoming permanently confined to markets instead of maturing into money.