The moment SoftBank announced Yossi Cohen as its strategic advisor for AI investments, the crypto world should have felt a chill. Not because of the man himself—though his résumé as former Mossad director is enough to raise eyebrows—but because of what this appointment signals: the final convergence of centralized capital, state intelligence, and artificial intelligence. For those of us who have spent years building decentralized, trustless systems, this is the clearest signal yet that the battle for the soul of AI is not a technical debate. It’s a geopolitical war, and the stakes are nothing less than the future of human agency.
Let me be blunt. I’ve been in this space since the 2017 ICO frenzy, when I launched three Telegram groups in one month for Ethereum projects from my Buenos Aires apartment. I’ve audited DeFi protocols that promised decentralization but had admin keys controlled by a single wallet. I’ve seen the pattern before: a system that claims to be open, transparent, and permissionless slowly morphs into a centralized hierarchy under the guise of efficiency. SoftBank’s move is the same pattern, but on a global scale. And if we in the crypto community do not react, we will wake up to find that the AI infrastructure of the future is owned by a cartel of the state, the corporation, and the intelligence agency.
Context: The SoftBank Machine
SoftBank is not your average venture capital firm. Under Masayoshi Son, it has become a force of nature, managing assets that at their peak exceeded $100 billion through the Vision Fund. Its crown jewel is Arm, the chip architecture company that powers nearly every smartphone and an increasing number of servers running AI workloads. SoftBank owns roughly 90% of Arm, giving it a chokehold on the foundational layer of the AI stack. When you combine that with the $100 billion war chest, a network of sovereign wealth fund partners (including Saudi Arabia’s PIF), and now a former Mossad chief, you have a machine that can shape the trajectory of AI development for decades.
Yossi Cohen’s role is officially described as a strategic advisor for AI investments. But let’s read between the lines. Cohen ran Mossad from 2016 to 2021, a period marked by aggressive cyber operations, targeted assassinations, and the expansion of Israel’s intelligence capabilities into the digital realm. His expertise is not in machine learning architectures or tokenomics. It’s in understanding how to wield information as a weapon, how to assess threats at a nation-state level, and how to build networks of trust in environments where trust is scarce. Those are precisely the skills that become invaluable when you are investing billions into AI startups that could one day control autonomous weapons, surveillance systems, or the infrastructure of global finance.
Core: The Three Levels of the SoftBank-Crypto Conflict
This appointment is not just about SoftBank. It’s about the collision of two worldviews. On one side, the centralized, permissioned, and state-aligned approach to AI. On the other, the decentralized, permissionless, and sovereign approach that crypto represents. Let me break down the three levels where this conflict will play out.
Level 1: AI Safety vs. AI Control
The AI safety community has long argued that powerful AI systems must be aligned with human values. But who defines those values? When an intelligence agency is involved in the investment process, the definition of “alignment” shifts. We don’t just need AI that is safe from accidental harm; we need AI that is safe from being turned into a surveillance tool. Cohen’s presence suggests that SoftBank is not just looking for the next OpenAI. It’s looking for AI that can be controlled, weaponized, and monetized within the existing power structures.
I’ve seen this before in the DeFi world. When I analyzed the governance token distribution of a supposedly decentralized lending protocol, I found that 80% of the voting power was held by two addresses that belonged to the founding team. The whitepaper talked about community governance, but the reality was a plutocracy. SoftBank’s AI investments will likely follow the same pattern: the technology will be open-source in name, but the key decisions—about model weights, training data, and deployment—will be controlled by a small group of people with intelligence backgrounds.
Level 2: The Custody of AI Models
In crypto, we talk about self-custody as a fundamental right. The same principle applies to AI. If you do not control the model that makes decisions about your life, you are not free. SoftBank’s model of AI investment is essentially a custody model: they provide capital, infrastructure, and now intelligence oversight, but the actual ownership of the AI remains with a centralized entity. This is the antithesis of what decentralized AI should be.
I founded a project called “Verifiable Minds” in 2026, which aimed to create a decentralized identity layer for AI agents. The idea was that any AI agent could prove its identity and actions using zero-knowledge proofs, without relying on a central authority. We were too early, and the project struggled with the technical complexity of zk-proofs for large models. But the principle remains: just as we need non-custodial wallets for our money, we need non-custodial AI for our decisions. SoftBank’s move makes it clear that the centralized forces are not going to voluntarily give up control.
Level 3: The Geopolitics of AI Infrastructure
Arm is the lynchpin. If you control the instruction set architecture, you control what software can run on the hardware. SoftBank’s ownership of Arm gives it the ability to influence the entire AI stack, from the chip to the cloud. Now, with Cohen, they have the ability to assess geopolitical risks and tailor their infrastructure investments accordingly. This is a double-edged sword. On one hand, it could lead to more stable and secure AI infrastructure. On the other hand, it creates a single point of failure that could be exploited by state actors.
I remember the 2022 bear market, when I spent months auditing the smart contracts of failed protocols. I discovered that many of the collapses were not due to code bugs, but to centralized decision-making that ignored the on-chain data. The same thing will happen with AI. If SoftBank becomes the dominant investor in AI infrastructure, any political or economic pressure on SoftBank could cascade into the entire AI ecosystem. We need alternatives, and those alternatives must be built on decentralized principles.
Contrarian: The Case for the Intelligence Community in AI
Now, let me play devil’s advocate. Some might argue that Cohen’s appointment is actually a good thing for AI safety. After all, the intelligence community has decades of experience in threat modeling, counterintelligence, and secure communication. They understand the risks of adversarial attacks on AI systems better than any academic. In a world where nation-states are already using AI for cyber warfare, having someone who thinks like an adversary could be an asset.
I’ve seen this argument before in the crypto space. Some people argue that institutional investors bring stability and legitimacy to a volatile market. They point to the ETF approvals in 2024 as a sign that crypto has matured. But I’ve always countered that institutional adoption comes with strings attached. The same is true here. Yes, Cohen could help SoftBank avoid investing in AI projects that are vulnerable to state-level attacks. But the price is that the AI itself becomes a tool of state power.
There is also a pragmatic argument: SoftBank’s capital and Arm’s architecture are so dominant that any attempt to build a decentralized AI alternative will have to interoperate with them. We cannot ignore the realities of the market. The question is whether we can build systems that are resistant to the centralization pressures that SoftBank represents. That is the challenge that every crypto founder working on AI should be thinking about.
Takeaway: The Choice is Ours
SoftBank’s decision to hire Yossi Cohen is not a minor HR move. It is a declaration of intent. It says that the future of AI will be built by a coalition of capital, intelligence, and technology, and that the rest of us will be consumers of that future, not participants. But we have a choice. We can either accept that narrative and build our projects within the SoftBank ecosystem, or we can build the infrastructure for a truly decentralized AI, one that is owned by the community, verified by the network, and controlled by no single entity.
I’ve been building in this space for almost a decade. I’ve seen the rise and fall of ICOs, the DeFi summer, the NFT mania, and the ETF era. Each time, the promise of decentralization has been tested by the forces of centralization. Each time, we have adapted. The AI wave is no different. The tools are different, but the principles are the same: trust no one, verify everything, and build systems that cannot be captured.
We don’t just build technology; we build trust. That trust is eroded when a former intelligence chief sits at the table of the world’s largest AI investor. But it can be restored when we build alternatives that are transparent, permissionless, and sovereign. Freedom isn’t just a feature; it’s the foundation. And our worth as a community isn’t measured by our market cap, but by our shared vision of a future where AI serves humanity, not the other way around.
I’ll leave you with this: the next time you hear about a startup that claims to be building decentralized AI, ask yourself who controls the hardware, who controls the training data, and who controls the model. If the answer is a single entity, no matter how benevolent, then it’s not decentralized. It’s just another layer of the machine. And the machine has just hired its most dangerous operator yet. The question is: will you be part of the machine, or part of the resistance?
