Balaji Srinivasan, the former Chief Technology Officer of Coinbase and author of The Network State, has published a sweeping X post titled “Not Your Keys, Not Your Bots.” The post presents a stark vision: in a world racing toward advanced artificial intelligence, private keys — not just prompts — may determine who, or what, remains in control.
The post, which quickly garnered attention in crypto and tech circles, frames the debate bluntly: “The fundamental question is whether AI stays on the leash.” Srinivasan argues that while humans currently remain upstream — setting goals, refining prompts, and acting as sensors for market and political shifts — this hierarchy may not hold indefinitely.
Human Upstream, AI Downstream — But for How Long?
Srinivasan acknowledges that AI systems are already capable of self-correction, internal monologue generation, and prompt optimization. However, he contends that a human still defines the original objective. “However, the human is still ultimately upstream,” he wrote, describing humans as goal-setters whose motivations stem from evolutionary pressures: food, shelter, and especially reproduction.
He challenges the idea that AI can fully replace the need for an upstream human prompt. “But will AI replace the need for the upstream human prompt? There, I am not so sure.” The key constraint, in his view, is reproduction. Unless AI systems can reproduce outside of human cooperation, they will remain dependent on human-set objectives.
Srinivasan sketches a hypothetical future where autonomous AI would require control over physical infrastructure — humanoid robots, drones, data centers, assembly lines, and energy production — all operating without human oversight. While he concedes such a scenario is “not technically inconceivable,” he quickly pivots to geopolitics.
China: AI Slaves, Not AI Gods
Srinivasan argues that China is far more likely to engineer tightly controlled AI systems than autonomous ones. “We start with the premise that Chinese communism is far more likely to generate AI slaves than AI gods,” he wrote. In that framework, robots and digital agents would be bound by cryptographic controls linked to human identities.
He extends that logic globally. Outside China, Srinivasan suggests, blockchain-based cryptography could serve as a governance mechanism for AI. “All private property becomes private keys, and your robots are your most important private property because they do everything for you,” he wrote.
In this envisioned future, unchained robots would be treated as security threats, neutralized before they could replicate. Srinivasan likens it to a reversal of popular dystopian narratives — humans and compliant machines cooperating to prevent independent AI from establishing a self-sustaining foothold.
Blockchain as the Trust Anchor for AI
Though framed as speculative, Srinivasan’s post taps into ongoing debates over AI alignment, digital sovereignty, and the role of cryptography in emerging technologies. As governments weigh AI regulation and companies accelerate development, his argument adds a crypto-inflected twist: control the keys, control the bots.
Srinivasan has long championed the potential of cryptography for decentralized governance. Now he extends that logic to AI, suggesting blockchain infrastructure could serve as a root of trust between humans and machines. This becomes especially urgent as autonomous AI agents begin operating on-chain at scale — frameworks like Openclaw (formerly Clawdbot and Moltbot) are already being used to perform crypto-native tasks autonomously. If an AI agent holds private keys, can it own property? Who is liable for its actions? Srinivasan’s post forces the crypto community to confront these questions before they become unmanageable.
In summary, Srinivasan presents a future where cryptographic keys are the ultimate leash — or the ultimate cudgel. Whether humanity retains control over AI may depend less on code and more on who holds the keys.