Vitalik Buterin outlines private AI security model

Ethereum co-founder Vitalik Buterin has outlined a private, security-focused artificial intelligence setup that runs entirely on local hardware to minimise data exposure.

Buterin said he avoids cloud-based AI tools due to privacy risks, instead using an open-source model configured with custom safeguards to control how AI interacts with external systems.

“The new two-factor authentication is the human and the LLM,”

Said Buterin.

His system includes a messaging daemon that allows AI to read messages but blocks outbound communication or crypto transactions unless manually approved by a human.

The approach is partly driven by security concerns, with Buterin citing research showing around 15% of tools built for the OpenClaw platform contained malicious instructions.

He also recommended that Ethereum wallet developers adopt similar safeguards, including transaction limits and human confirmation for higher-value transfers.

Buterin said the model reflects his broader philosophy of prioritising privacy and decentralisation, warning that reliance on cloud AI could reverse recent gains in data security.

At the time of reporting, Ethereum price was $2,055.27.