Vitalik Buterin linked DeepSeek V4 to the future of Ethereum privacy, outlining a roadmap that integrates native AI models into Ethereum’s access layer. The Ethereum co-founder specifically points out that there is significant overlap between the Ethereum access layer of CROPS and CROPS AI.
Buterin presented the concept of CROPS AI (Censorship-Resistant, Open Source, Private and Secure Artificial Intelligence) at the ETH Mumbai conference on March 12, where he discussed the reasons why AI could become the next major security risk for cryptocurrencies. He said that artificial intelligence has become powerful enough to manage wallets and interact with blockchains, but noted that the current ecosystem was not designed with privacy and security in mind.
Buterin believes that if AI agents are going to control cryptocurrencies, they need to be designed very differently. He says this reflects how far AI models have come.
According to Buterin, most people assume that the AI models running locally on their devices are private. But he confirms that this assumption is wrong.
The Ethereum chief points to the current state of native AI tools such as the Qwen 3.5 chain, proxy frameworks that run natively, and a growing body of open source software. He points out that while these models may appear to be autonomous on the surface, most call on OpenAI or Anthropic’s APIs whenever they need to perform a task they can’t handle on their own.
Buterin says DeepSeek V4 is vital to achieving local private transactions
An update on the progress of the CROPS AI project he’s been following, Buterin He says DeepSeek V4 (with a 2-bit quantized version running on 90GB of memory) is vital to achieving locally processed private transactions. He notes that the CROPS Ethereum access layer overlaps with CROPS AI, including ZK-based remote paid LLM calls and private Ethereum RPC reads. He calls for more fine-tuned AI models on Ethereum to improve the security of smart contracts and the protocol’s code.
“Another thing I have in mind is that there is actually a lot of crossover between ‘CROPS Ethereum Access Layer’ and ‘CROPS AI’. For example, we want a ZK way to make (paid) calls to remote LLMs. But if we had this, it would also be useful for solving another problem: reading private RPCs in Ethereum.”
–Vitalik Buterinco-founder of Ethereum
The Ethereum co-founder points out that the relationship between DeepSeek V4 and Ethereum’s privacy goals centers around the concept of CROPS AI. He points out that users can query Ethereum data using native models like DeepSeek V4 without revealing their metadata, IP addresses, or wallet balances to central RPC providers. DeepSeek V4’s ability to run on local, self-hosted setups ensures that users rely on self-sovereign infrastructure rather than the company’s own cloud servers.
Buterin suggests combining private local LLM calls with Ethereum ZK payments
Buterin suggests Combining private local LLM calls with Ethereum ZK proofs, allowing users to process their blockchain interactions privately off-chain. This helps hide transaction links on-chain, he says, noting that DeepSeek V4’s low hardware requirements are key to this. However, the 2-bit quantum version of DeepSeek V4 can also run on high-end consumer workstations.
Buterin also points out that the newly released DeepSeek V4 is essential proof that this vision is applicable to devices today, not years later. Users running DeepSeek V4 locally can create a “crypto haven” where their financial intent never leaves their physical device until it is ready to be added to the public ledger.
As for next steps, Buterin urges users to pay attention to AMD’s DeepSeek V4 Flash optimization patches, which he sees as a key area for improvement. It also reminds users to make sure their machines have at least 96GB to 128GB of unified memory (for Mac) or VRAM (for PC) to handle the 90GB of quantum load.
This push is linked to a broader “Cypherpunk” revival in which artificial intelligence acts as a proxy for users. Buterin emphasizes that this effectively mixes orders, separates payments from users’ identities and makes remote AI accounts anonymous.
Buterin also points to warnings from the cybersecurity community, noting that AI running locally may ping OpenAI’s servers when they become confused. He points out that the mainstream open source AI ecosystem doesn’t care about differentiation, adding that most of these systems are optimized for capability rather than security.
Don’t just read cryptocurrency news. Understand that. Subscribe to our newsletter. It’s free.
