SoftBank appointed Yossi Cohen, former Mossad director, as an AI strategic advisor. The market interpreted this as a signal of deep commitment to AI security. I interpret it as a declaration of control over the open, decentralized nature of AI development.
Context: The SoftBank AI Machine
SoftBank is not a typical VC. It owns Arm, the architecture behind nearly every AI chip. It controls the Vision Fund, a $100B+ war chest. Its founder, Masayoshi Son, publicly preaches AGI within a decade. The firm has pivoted from spray-and-pray investments to a concentrated bet on AI infrastructure. Now, it adds a former intelligence chief to its strategic core.
Core: The Systematic Teardown
This appointment is not about technology. It is about power. The code does not lie; only the auditors do. But here, the auditor is a former spy. Let me break down the layers.
First, the ethical fault line. Cohen led Mossad during a period of aggressive overseas operations. His background is not AI—it is surveillance, covert action, and state-level manipulation. When a firm with SoftBank’s capital and chip assets hires such a figure, the implied message is clear: AI is being weaponized, and we want to be the ones holding the weapon. This will repel AI labs that value ethical boundaries. OpenAI and Anthropic will think twice before deepening ties with a firm that has a Mossad veteran in its strategy room.
Second, the compliance nightmare. SoftBank has deep exposure to China (Didi, ByteDance) and the Middle East (Saudi PIF). A former Israeli intelligence officer sitting on the advisory board will trigger immediate scrutiny from Chinese regulators and may unsettle Middle Eastern partners. The Saudi-Israel normalization is fragile; this appointment could push PIF to reconsider its LP commitment. The silence from SoftBank on this risk is the loudest admission of guilt.
Third, the competitive distortion. The real value of Cohen is not his advice on AI safety—it is his network. He can open doors to Israeli security tech startups that are off-limits to most investors. This gives SoftBank an unfair advantage in accessing dual-use AI technologies. The market will cheer this as a ‘unique deal flow,’ but it is a form of regulatory arbitrage. I do not guess; I verify. The data will show whether SoftBank’s portfolio shifts toward surveillance and defense AI. If it does, the narrative of ‘AI for good’ is dead.
Contrarian: What the Bulls Got Right
Let me be fair. The bulls argue that AI safety is a national security issue, and that state-level threat assessment is necessary for large-scale AI investments. They point out that SoftBank’s ability to vet startups for foreign influence and technical rigor could protect against catastrophic failures. That is partially true. The contrarian angle is that Cohen’s expertise could actually strengthen SoftBank’s due diligence. In a world where AI models can be backdoored at the chip level, having someone who understands intel tradecraft is not just a luxury—it is a necessity.
But the bulls miss the critical point: this is a concentration of power. When a single firm controls the architecture (Arm), the capital (Vision Fund), and the intelligence lens (Cohen), it becomes a bottleneck. The same network that can protect can also manipulate. The same intelligence that can vet can also co-opt. The risk is not that Cohen is incompetent—it is that he is too competent for the wrong reasons. Every transaction leaves a scar on the ledger. This one scars the entire AI ecosystem.
Takeaway: The Forward-Looking Question
SoftBank’s move is not about AI; it is about control. The question is not whether Cohen adds value, but at what cost to the open AI ecosystem. The ledger will show the transaction, but the scars will be permanent. Promises are encrypted; data is decrypted. And the data here tells a story of centralization, not innovation. The market cheered, but the on-chain evidence says otherwise. I will be watching the next Vision Fund raise. If the LPs are silent, the guilt is confirmed.
