We assumed AI would cure diseases. We forgot to ask who will own the cure.
Last week, an interview snippet from the CEO of Anthropic rippled through the crypto timeline: "AI will cure most diseases within a decade." The market barely blinked. Then the DeSci tokens pumped. The narrative was neat: big AI, big biotech, big returns. But as I sat in a basement coffee shop in Beijing, auditing the governance mechanics of a biotech DAO that had raised $40M in tokenized research grants, the claim felt hollow. Not because it's impossible—but because the claim itself is a ghost in the machine. It signals a future where the cure is centralized, proprietary, and governed by a handful of corporate boards. And that, to me, is the disease we haven't diagnosed.
The code is law, but the humans are the bug.
Context: The Promise and the Void
Anthropic’s CEO, Dario Amodei, is no stranger to grand visions. In his 2024 essay "Machines of Loving Grace," he argued that advanced AI could compress a century of biomedical progress into five to ten years. The technology stack is real: large language models generating novel protein structures, generative models for molecular design, and agentic research automation that can run thousands of virtual experiments overnight. But the article that triggered this analysis—a Crypto Briefing piece—offered no technical details, no pipeline milestones, no clinical trial data. It was a signal. A PR signal. A signal that the AI industry is now competing for the soul of the life sciences, and that the blockchain industry, via DeSci (Decentralized Science), is being positioned as the financial layer for that future.
I have seen this playbook before. In 2020, during the DeFi Summer, I audited Curve Finance’s governance mechanics, analyzing over 400,000 lines of simulation data. The pattern was the same: a visionary promise—decentralized stablecoin trading—followed by a capital-weighted voting system that concentrated power among whales. The outcome was not a democratic utopia but a oligarchy of liquidity providers. Today, the same pattern is emerging in AI-driven biotech. The promise is a cure for all; the reality is a patent war over the few molecules that succeed.
We built a kingdom of ghosts in the machine.
Core: The Governance Architecture of the Cure
If AI is to cure most diseases, the question is not whether the algorithms work—it is who controls the data, who validates the models, and who decides which diseases are worth curing. The current centralized model, where a single company like Anthropic or Google DeepMind holds the keys to the most advanced protein models, is a recipe for monopolistic control. Based on my experience designing a quadratic voting mechanism for a $5M DAO treasury in 2024, I have seen how pluralistic representation can align incentives that pure profit cannot.
Let me be specific. The AI biotech pipeline can be broken into four stages: target discovery, molecule design, preclinical validation, and clinical trials. At each stage, there is a governance bottleneck.
- Target Discovery: Data is the new oil, but it is siloed. Academic institutions, hospitals, and biobanks hold vast genomic and phenotypic datasets. A centralized AI company can license this data, but the terms are opaque, and the communities that contributed the data rarely benefit from the resulting therapies. Decentralized data marketplaces, using zero-knowledge proofs and tokenized incentives, could allow patients to contribute their data to a DAO and receive royalties when a drug derived from that data reaches market. This is not a fantasy; projects like Genomes.io and Data Lake are already building these primitives. But the governance of those data pools—who sets the price, who approves the use, who audits the privacy—is still in its infancy.
- Molecule Design: The best models today are proprietary. AlphaFold is open in parts, but the most advanced versions are behind Google’s walls. A decentralized alternative, like the open-source protein language models from Meta’s ESM team, exists, but lacks the compute resources and curated datasets to compete. The governance challenge here is not just technical but economic: how do we fund the training of a truly open, community-owned foundation model for molecular biology? One approach is a DAO that issues tokens to researchers who contribute validation data, and uses those tokens to fund compute on a decentralized cloud. I have seen this model work in small-scale DeSci projects, but the capital required for a frontier model is in the hundreds of millions. The market is not ready for that level of patient capital.
- Preclinical and Clinical Trials: This is the valley of death. AI can reduce the number of failed experiments, but it cannot replace a human trial. The cost of a Phase III trial can exceed $1 billion, and the average drug takes 10-15 years to reach patients. The promise of AI is to compress the discovery phase, but the clinical phase remains the bottleneck. Here, blockchain-based trial registries, with on-chain audit trails of every data point, could reduce fraud and increase transparency. But the regulatory framework—FDA, EMA, and others—is not designed for decentralized governance. The question is not whether we can build a DAO to run a trial, but whether the FDA will recognize a trial governed by a token-weighted consensus.
Silence is the only consensus that never forks.
Contrarian: The Blind Spots of the Techno-Optimist
Now, the contrarian angle. The most vocal advocates of AI for biotech are also the most vocal advocates of AI safety. Anthropic itself is built on a safety-first ethos. But the contradiction is glaring: the same company that warns about catastrophic risks from AI is also promising a cure for all diseases within a decade. This is not hypocrisy; it is a strategic narrative pivot. The promise of a cure is the carrot that justifies the stick of centralization. If AI is to cure diseases, we need vast compute, vast data, and vast trust. Centralized companies are best positioned to deliver that—or so the argument goes.
But the blind spot is this: centralized control over the means of healing is itself a systemic risk. Consider the scenario where an AI model discovers a new class of antibiotics. The company that owns the model patents the family of molecules, sets the price, and decides which countries get access. The result is not a cure for all, but a cure for the wealthy. The blockchain community, with its ethos of permissionless access and transparent governance, offers a counter-narrative. But the DeSci movement is still small, fragmented, and mostly driven by speculation rather than science.
I have seen this before. In 2022, after the collapse of FTX and Terra, I spent six months in isolation, writing a private journal called "The Ethics of Ruin." The lesson I learned was that technology alone cannot fix broken incentives. The FTX collapse was not a failure of code; it was a failure of governance. The same will be true for AI biotech. The most elegant algorithm in the world cannot cure a disease if the governance system that distributes the cure is corrupt.
Intuition sees the pattern before the ledger does.
Takeaway: The Governance of the Cure
So where does this leave us? The Anthropic CEO’s statement is not a technical prediction; it is a political one. It is a vision of a future where a small number of AI labs control the fundamental tools of healing. The blockchain community, with its obsession with decentralization, has a role to play—but only if it moves beyond tokenizing speculation and builds the governance infrastructure for a new kind of science.
We need DAOs that can fund truly open AI models for biology. We need data cooperatives that give patients ownership of their own biomarkers. We need clinical trial protocols that are auditable on-chain, and regulatory frameworks that recognize decentralized governance as a legitimate form of oversight. These are not easy problems. They require years of work, not just a tweet.
To govern the future, we must debug the present.
The cure is not in the algorithm. It is in the consensus. If we build the governance first, the biology will follow. Otherwise, we will have a kingdom of ghosts in the machine—a system that promises to heal but delivers only more of the same: concentration, exclusion, and the quiet tragedy of unfulfilled potential.
I am Andrew Williams, a DAO Governance Architect based in Beijing. I have seen the code. I have seen the humans. The bug is not in the model. It is in the system we choose to build around it.