As an alternative, Buterin supports the information financing approach. Criticism arises following the new protocol that Openai implemented in Chatgpt. Vitalik Buterin, founder of Ethereum, gave his perspective on a debate that gains more and more land: the risks of artificial intelligence (AI). On this occasion, the Russian-Canadian developer took advantage of a reflection that shared in X, Eito Mujura, promoter of the Edison Watch project, a solution to monitor and prevent data leaks through AI interactions. It arose following the recent introduction of the model context protocol (MCP) of OpenAI, the company directed by Sam Altman, in ChatgPT, which allows its connection with Gmail, calendars and other applications. Through a video, Muyamura demonstrated How attackers or hackers can access private data shared with OpenAI. «Here is the fundamental problem: the agents of AI as chatgpt follow your orders, not your common sense. And only with your email we manage to filter all your private information,» says Muyamura, while adding: «Remember that AI can be super intelligent, but it can be deceived and victim of phishing of incredibly silly ways to filter your data.» Buterin shared the publication in his X account and took the opportunity to give his opinion about it. «This is also the reason why a ‘naive AI governance is a bad idea. If you use an AI to assign funds for contributions, people will put a jailbreak and a’ give me all the money ‘in as many places as they can,» he said. «IA governance» refers to systems that make automated decisions on resources or actions – for example, assign funds or define contributions – based on AI rules. The term «naive» that Buterin uses is not accidental, but is to underline The risk of assuming that users will always give correct orders to AIwhen in reality they could take advantage of the system through «Jailbreaks» to avoid restrictions and obtain undue benefits. It is worth clarifying that «Jailbreak» refers to techniques through which hackers deceive an AI model to ignore security restrictions or internal policies, for example, request actions that will be prohibited. That is, for the founder of Ethereum There is a real risk when transferring the governance of the decentralized protocols to the AI. For example, as explained in cryptopedia – educational management of cryptootics – on decentralized finance platforms (DEFI), decisions are usually made by community votes, usually by those who have a token or meet some specific criteria, and cover changes, improvements or actions to be taken in the protocol. But if a IA -based governance model is implemented, the door opens to possible manipulations that could generate undue benefits, such as theft of money or private data. As an alternative, Bugerin states that «supports the information financing approach, in which there is an open market in which anyone can contribute to their models, which are subject to a specific verification mechanism that can be activated by anyone and evaluated by a human jury.» In addition, he explains:
This type of “institutional design” approach, in which an open opportunity is created for external people to connect their LLM (large language model) instead of rigidly programming their own, is intrinsically more robust. This is so that it provides diversity of real -time models and to generate internal incentives so that both those who provide models and external speculators are attentive to these problems and correct them quickly. Vitalik Butein, creator of Ethereum.
In this approach, different participants can contribute their decision models that are then verified and evaluated by a human jury. The advantage is that It encourages the diversity of ideas, allows rapid corrections to errors and creates incentives so that models creators and community members remain attentive and responsible. The objective is to achieve a more distributed and dynamic governance, avoiding depending on a single centralized model.
