Vitalik Buterin Backs Kimi AI Limits on Cyber Capabilities

Vitalik Buterin Backs Kimi AI Limits on Cyber Capabilities


Ethereum co-founder Vitalik Buterin has backed efforts to limit advanced cyberattack capabilities in artificial intelligence (AI) models. He argued that restricting dangerous uses is preferable to limiting access based on users’ identities.

His comments came after Chinese AI startup Moonshot AI said it had removed cyber-related data from the training stages of its latest Kimi K3.1 Pro model to reduce its offensive cybersecurity capabilities while retaining general-purpose coding skills. 

Vitalik backs removing offensive cyber data from AI models, urging industry to curb AI-driven attack vectors.

Moonshot’s Kimi 3.1 Pro trims cyber training, sacrificing benchmark score but preserving open-weight accessibility for developers.

Debate highlights AI’s dual role in crypto: accelerating security tools while exposing blockchain to sophisticated vulnerabilities.

In a post on X on October 25, Buterin said he supported leaving advanced offensive cybersecurity capabilities out of the model’s training data. He argued that this approach could give governments, developers and security teams more time to prepare for increasingly powerful AI-driven attacks.

“I am okay with Kimi 3.1 deliberately leaving out top-end offensive cyber capabilities from their training data,” Buterin wrote.

He added that restricting specific use cases was “MUCH less bad than discriminating against users.” His comments draw a distinction between limiting a model’s ability to perform potentially harmful tasks and limiting who can access AI tools. 

Buterin Calls for Stronger Software Security

Buterin said the time before more powerful AI cyberattacks become widespread should be used to strengthen existing systems and improve software security. He highlighted the need to secure systems, review older code, and apply formal verification to new software. Formal verification uses mathematical methods to check whether a program meets specific requirements.

Advanced AI models can write code, use digital tools, and complete complex tasks. These abilities could help developers find software bugs faster, but they could also help attackers discover vulnerabilities and automate parts of cyberattacks.

The risk is particularly relevant to the cryptocurrency industry, where blockchain networks, smart contracts, and digital asset platforms depend on secure software. Vulnerabilities in wallets, smart contracts, or exchange systems can expose users to financial losses.

However, Buterin did not identify a specific attack against a crypto platform or claim that Kimi K3.1 Pro had been used to exploit blockchain vulnerabilities.

Moonshot Seeks to Limit Kimi’s Offensive Cyber Capabilities

Moonshot AI introduced Kimi K3.1 Pro as an open-weight model in the three-trillion-parameter class. In its official technical announcement, the company said it removed cyber-related data from the model’s training corpus and reinforcement-learning environments while retaining its general coding abilities.

As per the announcement, the modified model scored 81.0% on an internal cybersecurity benchmark, compared with 90.5% for a similar version without the risk-reduction measures. The test involved finding and exploiting software vulnerabilities in a controlled environment. 

Both scores were reported by Moonshot and have not been independently verified. The comparison suggests the measures reduced performance on that particular test; it does not establish that the model is unable to carry out offensive cyber tasks.

Moonshot said it did not want to release model weights that would push offensive cyber capabilities further. It had originally planned to release the model’s full weights on November 4, 2026.

Open-weight models allow developers to download and run AI systems on their own infrastructure. While this can help organizations test software and improve security, it also makes it harder for the original developer to monitor use or enforce restrictions after release.

The decision highlights a challenge for AI developers: keeping useful coding capabilities while reducing the risk of misuse.

AI Brings New Risks to Crypto Security

Buterin has warned that AI could create new risks for cryptocurrency security while also helping developers build safer software.

On October 8, he warned that AI-driven advances in mathematics could weaken cryptographic systems, including lattice-based cryptography. He advised users not to rush into wallet migrations, as poorly managed transfers can create additional risks. He also discussed Ethereum’s move toward hash-based cryptography to reduce reliance on certain mathematical systems that could face future security challenges.

At the same time, Buterin believes AI could help improve software security. In May 2026, he highlighted AI-assisted formal verification as a way to help developers identify code errors before deployment. Formal verification uses mathematical methods to check whether software meets specific requirements. However, it cannot prevent every security failure. 

In February 2024, Buterin identified code bugs as one of Ethereum’s biggest technical risks and supported using AI to help verify code and detect vulnerabilities. He also warned that relying on AI without proper checks could be dangerous, particularly for high-value applications such as stablecoins and prediction markets.

As AI tools become more capable, they could help developers find bugs faster but may also miss vulnerabilities or produce incorrect results. AI agents with access to crypto wallets could introduce further risks if they misunderstand instructions or make unwanted transactions. Limiting their permissions and requiring human approval for sensitive actions can help protect users’ funds.

Buterin Warns Against Identity-Based AI Restrictions

Buterin also warned that a poor response to AI risks could lead to highly restricted models, widespread know-your-customer (KYC) checks and intrusive surveillance. In such a scenario, users could receive different levels of access based on their credentials or nationality.

His concern is the difference between limiting dangerous capabilities and restricting who can use AI. The first approach focuses on potentially harmful activities, while the second could prevent people from accessing useful tools regardless of their intended use.

Buterin presented this as a possible risk, not a confirmed policy direction. His comments highlight the challenge facing AI developers and governments as they try to reduce cybersecurity threats without imposing unnecessarily broad restrictions.

For the crypto industry, the priority is to strengthen software security, improve code verification and limit the permissions given to AI agents. The debate over Kimi K3.1 Pro also raises a wider question: how can developers reduce the risk of AI-powered attacks while keeping useful technology accessible?

Also Read: Ethereum Delays Oct 26 Glamsterdam Hoodi Upgrade Over Security Risks


Disclaimer: The information researched and reported by The Crypto Times is for informational purposes only and is not a substitute for professional financial advice. Investing in crypto assets involves significant risk due to market volatility. Always Do Your Own Research (DYOR) and consult with a qualified Financial Advisor before making any investment decisions.





Leave a Reply

Your email address will not be published. Required fields are marked *