Skip to content
- Buterin says CROPS AI supports local AI, secure messaging, and private Ethereum access.
- CROPS AI tools include Telegram support, VoxTerm transcription and Lucebox Hub updates.
- Zero-knowledge API calls could reduce data exposure in Ethereum RPC reads for users.
Ethereum co-founder Vitalik Buterin has outlined new updates for CROPS AI, a privacy-focused framework for local artificial intelligence, secure messaging, and protected Ethereum access. He shared the update in a Thursday post on X, citing several open-source tools.
Buterin said the projects support self-sovereign AI systems. The work focuses on reducing reliance on centralized cloud platforms. It also connects AI privacy tools with Ethereum infrastructure.
CROPS AI Tools Focus on Safer Local AI Use
The CROPS AI update builds on Buterin’s April essay about private LLM setups. In that post, he warned about weak safeguards in many AI systems. He cited prompt injection, data leaks, and unrestricted agent permissions.
One project in the latest update was messaging-daemon. The tool now includes alpha Telegram support. It lets AI systems use messaging apps while requiring manual approval for risky actions.
Buterin also mentioned VoxTerm, a local AI transcription tool. It operates without any third-party servers. The project aligns with the CROPS AI vision of user information remaining closer to devices.
Lucebox Hub was another tool. It is designed to improve performance for dense AI models, including Qwen 27B. Buterin also mentioned a 2-bit quantized DeepSeek v4 model, which can be deployed within 90 GB of memory.
This is still a variation in performance, depending on hardware, he added. Buterin says that the model is much more efficient when run on Apple hardware than on AMD hardware. He did not present the gap as fully resolved.
Also Read: Mastercard Secures BitLicense to Expand Stablecoin Payment Infrastructure
Ethereum privacy was also addressed during the CROPS AI discussion. Zero-knowledge API calls to remote AI models could assist with Ethereum RPC reads, said Buterin. This could minimize request data exposure when customers use blockchain-based services.
Buterin Connects CROPS AI to Ethereum RPC Privacy
The privacy of RPC on Ethereum is still relevant for the users. In many cases service providers can know the data that users request. Buterin’s remarks indicate that comparable privacy techniques might lead to both AI accessibility and blockchain queries.
Buterin also stated that he wanted some AI models based on Ethereum. He referred to Leanstral, a model designed for Lean programming tasks. He added that there should also be fine-tuned models related to development and security for Ethereum.
Local-first models, sandboxing, offline knowledge storage, and human approval layers are incorporated into the CROPS AI approach. Buterin previously listed tools such as Llama Server and Bubblewrap. He also mentioned the permissionless connections between AI agents and Ethereum wallets.
His broader point was not that AI should be avoided. Rather, he claimed that systems should have more privacy control. The design of the CROPS AI framework is built around user control.
In a recent update, Buterin connects AI security to blockchain access. It shows his continued focus on tools that limit surveillance and automated abuse. CROPS AI now serves as a wider label for that privacy-first direction.
Also Read: Bitwise Hyperliquid ETF Hits Record Inflows as Bitcoin ETFs See Outflows
