Ethereum co-founder Vitalik Buterin has raised concerns that the centralization of power over artificial intelligence could pose a greater risk to humanity than the technology’s own intelligence.
Power concentration poses core AI threat
Buterin argued that public debate around AI too often centers on whether advancing machine intelligence itself will endanger society, rather than examining the implications if control of these systems falls into the hands of a few entities.
He stated that if AI eventually outperforms humans at all critical tasks, humanity’s collective bargaining power could fall to zero, fundamentally shifting the balance of power worldwide. Buterin pointed out that scenarios in which a small number of AI laboratories dominate development and deployment should be a source of widespread concern.
He highlighted that, in any context outside the tech sector, a few organizations holding disproportionate power would be considered dangerous and unacceptable.
Buterin warned against complacency, noting that assuming a smooth and equitable transition toward advanced AI is, in itself, a risky and underexamined belief. He questioned why calls for open discussion and potential acceleration of AI receive greater skepticism compared to assumptions that centralized governance will safely manage future risks.
Open-source mandates seek to prevent dominance
The Ethereum co-founder referenced updated provisions in major AI roadmaps mandating open-source access in AI development as an important step to reducing the risk of monopolization. By allowing broader access to foundational AI models and code, these changes are intended to prevent a handful of companies from controlling transformative technologies.
Buterin also pointed to the concept of “mutually assured compute destruction,” a proposed collective safeguard which would enable multiple parties to agree in advance to halt large-scale AI computation under specific conditions. This mechanism is designed to balance intervention powers and prevent any single actor from disenfranchising others.
He emphasized that such systems are preferable to frameworks allowing select groups to exclude opponents or dissenting voices from key governance decisions.
Mini dictionary: Mutually assured compute destruction is a governance mechanism where several actors hold the power to pause or slow global AI operations collectively, reducing individual dominance and enforcing checks on potentially dangerous technology escalation.
Buterin backs decentralized defense against AI risks
Buterin outlined his support for d/acc, a decentralized defensive strategy focused on countering AI-related existential threats. This approach gives priority to formal verification, advanced cryptography, secure open hardware, pandemic readiness, and food security.
Additional components include improving public epistemics—systems that support reliable shared facts—and preparing for risks that may surface regardless of whether AI develops as quickly as predicted.
He suggested that pre-agreed triggers, such as the onset of severe pandemics or unemployment rising above 25%, could mediate disputes between those favoring rapid AI progress and those urging caution. Once these conditions are met or fail to occur, legitimacy may shift toward whichever group’s predictions prove correct, making debates more fair and inclusive.
Buterin expressed that direct involvement in these decisions should extend beyond governments and executives, inviting wider community participation through public platforms and social media.
He acknowledged that every proposal, no matter how well-designed, contains elements critics might label as unrealistic. Yet, he stressed the importance of engaging openly with efforts to address the unique challenges that AI presents to society.
Buterin ultimately maintained that while a perfect governance scheme does not yet exist, he welcomes ongoing proposals—especially those that aim to reduce concentrated control over AI development.
