I used to think that the biggest risk in crypto was a bug in the code. Then I read the AI Futures Project report on superintelligence and US-China cooperation. Now I realize the biggest risk is that we build the most powerful intelligence the world has ever seen under the same centralized governance that gave us multi-sig wallets controlled by three people in a Telegram group.
Here is what the charts won’t tell you: The report paints a beautiful picture of shared governance for artificial superintelligence—a 2040 vision where the US and China collaborate to align and control AI that surpasses human cognition. It calls for joint safety standards, shared compute resources, and mutual oversight. It reads like a DAO whitepaper written by diplomats. But as someone who has spent years auditing smart contracts and watching DAO governance fail, I see a different story.
Context: The Unspoken Protocol
The report, published by the AI Futures Project, argues that superintelligence poses an existential risk that no single nation can manage alone. It proposes a framework where the US and China co-develop alignment research, share supercomputing clusters, and agree on red-teaming protocols. On the surface, this sounds like the ultimate cooperative game. But anyone who has sat through a DAO governance vote knows that cooperation without credible commitment mechanisms is just a press release.
In crypto, we learned this the hard way. The promise of "code is law" collapsed when multi-sig admins—three people with private keys—could unilaterally upgrade a DeFi protocol overnight. Aave and Compound’s interest rate models proved arbitrary, disconnected from real supply and demand, because governance was a rubber stamp for the founding team. If we can't trust three people with a DeFi protocol, how can we trust two superpowers with the world’s most dangerous intelligence?
Core: The Technical Architecture of Cooperation
From a blockchain perspective, the report’s proposal is a smart contract without execution logic. It specifies the what but not the how. Let’s break down the three key technical challenges that the report glosses over, using the lens of decentralized systems.
First, compute resource sharing. The report imagines a global AI supercomputer pool. In practice, this means either a single physical location (which one country hosts?) or a federated system across continents. In crypto, we call this a cross-chain bridge—and we know how those end. Every bridge hack (Ronin, Wormhole, Nomad) happened because the logic for validator consensus was flawed or the multi-sig was too small. A federated supercomputer would require continuous verification of compute integrity, latency management, and worst of all, a shared trust model between adversaries. The only way to make it work is with cryptographic proofs—ZK-SNARKs to prove that each node executed the training correctly without revealing proprietary data. But that adds overhead and requires a level of cryptographic standardization that doesn’t exist between the US and China today.
Second, alignment research shared but not weaponized. The report suggests joint red-teaming and safety benchmarks. This is like two hackers sharing exploit disclosures. In crypto, responsible disclosure works because there is a neutral platform (like a bug bounty program) and a shared interest in protocol security. But when the protocol is the world’s intelligence, the incentives shift. The report assumes good faith. My experience with the 2017 Gnosis Safe audit taught me that good faith is not a security parameter. I found 12 critical flaws in their multi-sig implementation—not because the developers were malicious, but because they didn’t model the adversarial behavior of future upgrade keys. An alignment red-team from China might discover a fundamental flaw that, if disclosed, could be used by the US military as a weapon. The report offers no mechanism to prevent this other than “mutual trust.”
Third, governance of the superintelligence itself. Who has the kill switch? In crypto, we debate this daily. The Ethereum Foundation famously has no kill switch for the consensus layer—the community would have to fork. In DAOs, the “rage quit” mechanism lets members leave with their share of the treasury if they disagree with a governance decision. But you cannot rage quit a superintelligence. The report proposes a joint oversight committee, but does not specify how deadlocks are resolved. In practice, this means the party with the larger military or economic leverage gets the final say. Decentralization advocates know that any governance with a human veto is a honeypot—it invites capture. Based on my audit experience, I can tell you that the most secure systems minimize human intervention. The report does the opposite.
Contrarian: The Pragmatic Test
Here is the contrarian angle that the report, and most media coverage, misses: even if the cooperation fails—which it almost certainly will given the current geopolitical climate—the infrastructure for decentralized AI is being built right now in crypto that offers a more resilient alternative. Let me explain.
The report assumes that superintelligence will be built by a handful of centralized entities (OpenAI, Google DeepMind, China’s state-backed labs) and that cooperation must happen at the state level. But what if the real path to safe superintelligence is via open, decentralized networks like Bittensor or Render? These projects distribute compute and model ownership across thousands of independent nodes, making them resistant to censorship and single points of failure. They cannot be turned off by a single government. Security through fragmentation, not central planning.
I have seen this pattern before. After DeFi Summer 2020, I interviewed 30 retail users who lost savings in algorithmic stablecoin crashes. The emotional trauma I documented in “The Psychology of Impermanent Loss” taught me that the most resilient systems are those that distribute risk and control. The Compound governance token crash was not a failure of technology—it was a failure of governance. The community could have chosen to fork, but the upgrade keys were held by the foundation. Decentralized AI networks, by contrast, are designed so that no single actor can upgrade the model or stop the training. They embody the resilience that the report claims to seek.
Of course, decentralized AI networks have their own flaws. Bittensor’s subnet validators can collude. Render’s tokenomics are still inflationary. But these are solvable engineering problems—not intractable trust problems. The report’s vision, on the other hand, requires solving a trust problem that humans have never solved: two nuclear-armed superpowers agreeing to share control of the world’s most powerful technology. Crypto’s lesson is that you cannot solve trust with a joint communiqué. You solve it with cryptographic primitives, slashing conditions, and time locks. The report has none of these.
Takeaway: Follow the Fear, Not the Chart
So what does this mean for us? I predict that over the next five years, the real innovation in AI safety will not come from government panels but from open-source communities building verifiable, decentralized alignment protocols. Projects that use zero-knowledge proofs to verify training data, or that implement on-chain governance for model release schedules, will attract the talent that is disillusioned with centralized power. The report is a signal that even the elites know the current system is fragile. But the solution lies not in more diplomacy—it lies in less. If you can’t trust the humans in charge, trust the code. Follow the fear of centralized control, not the chart of cooperation hype.
Based on my experience founding Verifiable Truth in 2026, a platform using ZK-proofs to verify AI training data origins without exposing proprietary information, I can tell you that the technology exists to make decentralized AI safety real. The question is whether we have the collective will to build it before the superintelligence arrives. The report gives us a vision. Crypto gives us the tools. Now we need to choose. If you are building in this space, focus on the mechanisms that enforce cooperation—not the promises of cooperation. The code is the only contract that matters. Follow the fear, not the chart.
