Evan Kotsovinos left Google’s safety AI team to become Goldman Sachs’ head of artificial intelligence. The press release spun it as a strategic shift. I see it differently: a deliberate move to weaponize big tech's compliance machinery against the one sector that still resists central control—crypto.
Goldman Sachs called Bitcoin a ‘scam’ in 2018. Today it holds billions in crypto derivatives and runs a digital asset desk. Now it hires the man who built Google’s AI safety protocols. This is not about trading algorithms. It is about surveillance.
The context is critical. We are in a bear market where every crypto protocol is bleeding liquidity. Meanwhile, Wall Street is scooping up AI talent from the same companies that gave us censorship-friendly models. Kotsovinos spent years at Google ensuring large language models did not expose vulnerabilities or violate internal policies. He was the gatekeeper. Goldman wants that gatekeeper mindset applied to its own systems—and by extension, to the crypto assets it touches.
Here is the core insight, drawn from my own forensic work. Over the past four years, I have audited 45 smart contracts for pre-ICO startups and traced the ghost liquidity of three collapsed yield farms. The pattern is consistent: when a centralized entity deploys AI to manage or monitor decentralized systems, it always prioritizes compliance over permissionlessness. Goldman’s AI chief will not revolutionize trading. He will automate the detection of suspicious transactions, flag DeFi protocols that violate sanctions, and generate reports for regulators. The code whispered truth; the balance sheet lied. But now the balance sheet has its own AI.
Data supports this. Goldman’s 2023 operating costs were roughly $33 billion, with compliance and risk management accounting for an estimated $3 billion. A large language model that automates know-your-customer checks and anti-money laundering reviews could cut that number by 30%. That is nearly a billion dollars in savings. Kotsovinos’s background in safe AI deployment is the perfect match for this cost-cutting agenda.
But there is a darker layer. Every blockchain story ends in a forensic audit. With Kotsovinos in charge, Goldman will likely deploy its AI to scrutinize on-chain activity in real time. Consider their role as a custodian for Bitcoin ETFs. They hold the keys—or at least the paper representing them. Their AI could monitor wallets, flag privacy coins, and even recommend blocking transactions that originate from mixers. The smart contract does not care about your hopes. But a Goldman-trained model will care deeply about regulatory risk.
Now the contrarian angle, because no teardown is complete without acknowledging what the bulls got right. The hire could legitimize the intersection of AI and crypto. If Kotsovinos pushes Goldman to adopt transparent, explainable AI models that are auditable on-chain—something his Google background might enable—it could set a standard for institutional-grade DeFi. He might even advocate for open-source safety frameworks, as he did in parts of his Google tenure. That would be a win for verifiability, the very property crypto champions.
Furthermore, the competition for AI talent between Wall Street and Big Tech raises the floor for everyone. Crypto AI projects like Bittensor or Render Network might lose a few engineers to Goldman’s higher salaries, but they also gain credibility when a traditional finance giant validates the idea of decentralized AI compute. I traced the ghost liquidity back to its source during the Terra collapse. That source was centralization. If Kotsovinos brings some of Google’s distributed systems discipline to Goldman, it might ironically push the bank toward more decentralized infrastructure, not less.
But do not mistake this for optimism. The net effect is still a tightening noose around crypto’s neck. Goldman’s AI will not be used to improve Uniswap’s hooks or optimize L2 scaling. It will be used to filter which transactions are allowed, which wallets are trusted, and which protocols get institutional access. Silence in the logs is louder than the hack. If a DeFi application suddenly stops receiving Goldman’s custody business because an AI model flagged it as suspicious, there will be no alarm. Just a silent rejection.
The takeaway is simple. Goldman’s Google grab is not about innovation. It is about control. The same technology that makes AI powerful—pattern recognition, anomaly detection, automated enforcement—is now being aimed at the crypto ecosystem. Every token holder should ask: when the bank’s AI decides your transaction is too risky, who will you appeal to? There is no oracle for that.
The code still whispers truth. But Goldman just hired the person who decides which truths get heard.