Vitalik Buterin Supports AI Model Kimi K3.1 Pro's Hacking Limitations

Vitalik Buterin, co-founder of Ethereum, has expressed his support for Moonshot AI's recent decision to exclude advanced offensive hacking capabilities from its Kimi K3.1 Pro model. This move, he argues, is preferable to restricting user access to AI technologies.

In a post shared on October 26, Buterin highlighted the need for developers to have the necessary time to enhance system security and conduct formal verification of software before AI-driven cyberattacks become commonplace. He stated, "Discriminating against use cases is MUCH less bad than discriminating against users," reinforcing his belief in the importance of responsible AI deployment.

Moonshot AI, a Beijing-based startup, has reported that the filtered Kimi K3.1 Pro model successfully solved 81% of tasks on its internal exploit benchmark, compared to 90.5% for an unfiltered version. The company plans to release the model's open weights by November 4, allowing users to run it on their own hardware.

Buterin's concerns extend beyond technical capabilities; he warned that significant failures in AI security could lead to increased surveillance and identity checks, potentially creating a tiered system of access based on credentials and nationality. He emphasized the need for proactive measures to ensure equitable access to AI technologies.

Despite his support for the Kimi K3.1 Pro's limitations, Buterin acknowledged that other labs are likely to develop models with advanced hacking capabilities, suggesting that the race for AI safety is ongoing. He advocates for the implementation of robust security measures to stay ahead of potential threats.