The Spy Who Came in from the Code: What SoftBank's Mossad Hire Means for AI's Trust Architecture
When the former head of Israel's Mossad takes a seat at the table of the world's most aggressive AI investment fund, the message is clear: the future of AI is not just about code, but about control. Yossi Cohen, who led Israel's intelligence agency from 2016 to 2021, has been appointed as a strategic advisor to SoftBank, specifically to guide its AI investments. The announcement was brief, almost clinical—a press release that buried the lead in corporate jargon. But for anyone who has spent years watching the intersection of technology, power, and trust, this is a seismic event. It's not just another hire. It's a signal that SoftBank, already one of the largest forces in AI capital, is now embedding the logic of state surveillance into its investment thesis. And that should give every builder, every open-source contributor, and every believer in decentralized technology pause.
To understand why, we need to step back. SoftBank is not a typical venture capital firm. It is a capital aggregation machine, powered by the Vision Fund, which at its peak managed nearly $100 billion. Its founder, Masayoshi Son, has publicly declared that he believes AGI will arrive within the next decade, and that SoftBank's mission is to become the 'infrastructure provider for the AGI era.' The centerpiece of this strategy is Arm, the chip architecture company that SoftBank still controls after its partial IPO. Arm's designs are the foundation of nearly every AI chip in mobile devices and an increasing share of servers. With Arm, SoftBank owns a critical layer of the AI stack. Now, with Cohen, it is adding a layer of intelligence and security that goes far beyond technology.
The official narrative, of course, is that Cohen will help SoftBank navigate the 'geopolitical risks' and 'safety considerations' of AI investments. That sounds benign. But if you have ever audited a smart contract or worked in community governance, you know that the language of 'risk management' often masks a deeper shift in power. In my years running blockchain literacy circles at Zhejiang University and later doing on-chain analysis for DAOs, I learned that trust is not a resource you can delegate. It is something you build through transparency, verification, and shared ownership. The intelligence world operates on the opposite principle: trust is a weapon, a currency of secrecy, and a tool of control. The appointment of a former Mossad chief to the highest echelons of AI investment is a direct injection of that mindset into the heart of the AI development pipeline.
Let's examine the core of what this means. The most immediate impact is on the types of AI startups that will receive SoftBank's capital. Historically, SoftBank's AI investments have been broad—everything from autonomous driving (Cruise) to robotics (Boston Dynamics) to natural language processing (various). But Cohen's background signals a sharp pivot toward what I call 'security-first AI'—companies that build technologies for surveillance, cyber defense, adversarial testing, and military applications. Israel is a global hub for such technologies, often originating from units like 8200, the intelligence corps that has spawned countless startups. Cohen's network provides SoftBank with a privileged access point to these deals, a channel that no other fund can replicate. This is not just about money; it is about trust. In the intelligence community, deals are done based on personal relationships and shared background. Cohen can open doors that are closed to outsiders. That is a massive competitive advantage for SoftBank, and a massive concentration of power.
But the deeper issue is what this means for the ethical architecture of AI. I have spent years arguing that blockchain is not just a financial technology; it is a philosophy of trust. Smart contracts, transparent ledgers, community governance—these are tools for building systems where trust is distributed and verifiable. The SoftBank-Cohen partnership represents the exact opposite: a system where trust is centralized, hierarchical, and opaque. The same companies that receive SoftBank funding will be evaluated not just on their technical merit, but on their alignment with a security paradigm that is fundamentally at odds with principles of open source and decentralization. If you are a startup building decentralized AI models, or a privacy-preserving computation layer, you are now less likely to get SoftBank's money. Instead, the capital will flow to companies that sell to governments, that build surveillance tools, that treat AI as a weapon rather than a public good.
This is not a hypothetical. I have seen the dynamics of trust in blockchain communities firsthand. In 2022, during the bear market, I ran a weekly webinar series called 'DeFi for Humans' where I taught over 200 students how to secure their assets and understand smart contract risks. I helped dozens of people recover lost funds by walking them through code audits. What I learned is that the most resilient systems are those where every participant can verify the rules. That is the opposite of the intelligence model, where the rules are secret and the power is concentrated. Bringing Cohen into SoftBank is an implicit declaration that the future of AI will be governed by the same logic: a small group of trusted experts making decisions behind closed doors, with the backing of state-level surveillance networks.
Now, let me offer the contrarian angle—because I am not here to simply preach to the choir. There is a pragmatic argument that AI safety is so urgent that it requires the highest levels of adversarial expertise. Cohen's background in counter-terrorism and intelligence could indeed help SoftBank avoid catastrophic investments—for example, funding a company that inadvertently creates a dangerous AI vulnerability. The world has seen too many cases where technology was released without proper security testing. Maybe, just maybe, having a former spy on the team could prevent the next disaster. Some would argue that if we are serious about AI alignment, we should welcome any expertise that can help us navigate the risks, even if it comes from a background we find uncomfortable.
But that argument fails a critical test: it conflates safety with control. The intelligence community's definition of 'safety' is the preservation of existing power structures. In blockchain, we define safety differently: the ability of a network to resist capture by any single actor. There is a fundamental tension between these two definitions. Cohen's presence will nudge SoftBank's investments toward solutions that are 'safe' in the sense of being state-friendly, not in the sense of being robust against censorship or resistant to surveillance. The very technologies that could empower individuals—decentralized identity, privacy-preserving AI, open-source models—will be seen as risks, not opportunities. Trust isn't compiled, verified, and shared in a system like that. It is hoarded, weaponized, and used to maintain control.
I have seen this pattern before in the blockchain ecosystem. When I was auditing tokenomics for five open-source projects in 2017, I noticed that the most successful DAOs were not the ones with the most money, but the ones with the most transparent governance. The same principle applies to AI. The long-term health of the AI ecosystem depends on our ability to build systems that are open, auditable, and accountable. SoftBank's move, while perhaps rational from a short-term risk management perspective, is a step in the opposite direction. Code is only as strong as the trust it protects, and when that trust is placed in a former intelligence chief, the code itself becomes a tool of surveillance.
What does this mean for the rest of us? For developers building on blockchain, this is a wake-up call. The AI-crypto convergence is not just about merging two technologies; it is about two competing philosophies of trust. SoftBank's bet is that the future will be controlled by a few powerful actors who use AI as a force multiplier for their own agendas. Our bet, in the decentralized community, is that the future will be built by many, with transparent rules and shared ownership. The appointment of Yossi Cohen is a reminder that the other side is well-funded, well-connected, and willing to use any tool to maintain its dominance. Bridges aren't built by surveillance; they are built by trust. And trust is not something you can buy with a Venture Capital fund. It is something you earn, one open-source commit at a time.
So as we move forward, we need to be clear-eyed about the stakes. The AI arms race is accelerating, and the players are not just companies—they are states, intelligence agencies, and the blurred lines between them. The blockchain community has a unique opportunity to offer an alternative model: one where AI is not a weapon but a commons, where trust is not a secret but a protocol. We don't build the future by delegating our safety to spies. We build it by making the code so transparent that anyone can verify it, and the governance so inclusive that everyone has a voice. The next time you read about a new AI fund or a flashy appointment, ask yourself: whose trust is being protected, and whose is being sacrificed? The answer will determine the kind of world we leave for the next generation of builders.