His warning comes at a time when the conversation around AI agents running financial and organizational systems is heating up, but he insists that the risks far outweigh the appeal.
The concern is simple: AI systems can be tricked. Buterin argued that if such tools were tasked with managing funding decisions, bad actors would quickly find ways to game them. Rather than distributing resources fairly, the machines could be manipulated with jailbreak-style prompts designed to funnel money into the wrong hands.
This debate intensified after a discovery by Eito Miyamura, CEO of EdisonWatch, who recently exposed a vulnerability tied to ChatGPT’s latest upgrade. The update allowed the chatbot to connect seamlessly with everyday apps like Gmail, Notion, and Google Calendar. While the integration was pitched as a productivity booster, Miyamura demonstrated how it could be exploited to strip users of their private data.
The process is disturbingly straightforward. An attacker could send a calendar invite laced with a hidden jailbreak command. Even without accepting the invite, the victim would be at risk if they later asked ChatGPT to prepare their daily agenda. Once the AI reads the malicious entry, it can be hijacked, giving the attacker control to dig through personal emails and forward the contents elsewhere.
For Buterin, this illustrates exactly why AI should not be handed the keys to governance. Instead, he favors what he describes as an “info finance” system — a marketplace where competing AI models operate under constant scrutiny. In this design, independent reviewers and human juries, supported by language models, act as watchdogs. The presence of diverse models and a system of spot-checks ensures that flaws are uncovered quickly, while incentives keep developers honest.
The appeal of having a machine run complex institutions might be strong, Buterin admitted, but he stressed that the reality would likely be messy. Without built-in safeguards and human oversight, he believes attempts to crown AI as an autonomous governor could result in catastrophic losses rather than progress.
The information provided in this article is for educational purposes only and does not constitute financial, investment, or trading advice. Coindoo.com does not endorse or recommend any specific investment strategy or cryptocurrency. Always conduct your own research and consult with a licensed financial advisor before making any investment decisions.
The post Ethereum Co-Founder Says AI Rule Could End in Disaster appeared first on Coindoo.
Also read: Spur Protocol Daily Quiz Answer 15 September 2025: Earn Rewards
