Vitalik Buterin Endorses AI Model Without Top Hacking Skills

Vitalik Buterin, the Ethereum co-founder, has expressed support for Moonshot AI's decision to remove high-level hacking capabilities from its new Kimi K3.1 Pro model. He believes restricting the technology is better than restricting the user.

But this safety measure comes with a cost. Moonshot reports that the filtered version completed 81% of tasks in its exploit benchmark, while the unfiltered version succeeded at 90.5%.

Why Buterin Favors AI Restrictions Over User Screening

Buterin outlined his view, arguing that differentiating between uses is far less harmful than differentiating between users.

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

Discriminating against use cases is MUCH less bad than discriminating against users. And the world does need all the time it can get to prepare, harden, check old…

— vitalik.eth (@VitalikButerin) October 26, 2026

Moonshot, a Beijing-based AI startup, plans to release the model's open weights by November 4. After that, anyone can download and run it on their own hardware.

For that reason, the company says it removed cyber-related data from its training corpus and reinforcement learning environments. General coding tasks remained.

According to Vitalik Buterin, AI-powered attacks will eventually hit with full force. Until then, developers need time to harden systems, check legacy code, and verify new software.

The threat is already visible. Earlier this year, Binance Research found that AI exploits smart contracts roughly twice as well as it detects flaws.

Outcomes if the First Wave of AI Risk Goes Wrong

Buterin also warned about the alternative. If this first wave goes badly, he expects locked-down models, Know Your Customer (KYC) checks on everything, and intrusive surveillance.

In that scenario, AI access could depend on credentials and nationality. Some people would become second-, third-, or fourth-class users, he wrote.

Vitalik Buterin has long favored running open AI models on his own hardware, as his local AI laptop tests showed.

Still, he does not expect the filtering to hold forever. When a user asked whether other labs would build such capabilities anyway, he replied that they would.

In fact, by Moonshot's own count, DeepSeek's rival open model V4.1-Flash already scored 85.7% on the same benchmark.

For Vitalik Buterin, AI safety has therefore become a race for time rather than a permanent block. That race favors defenders who adopt his redundant crypto security safeguards before attackers catch up.