Jennifer Bryan, MD

⚡️ AI is beginning to separate legal authority from the human expertise that historically justified it. For centuries, organizations needed lawyers because understanding the rules was expensive. Interpreting contracts, regulations and liability required scarce human judgment, giving the legal profession enormous economic leverage. Now that expertise is moving directly into corporate workflows. Employees can consult systems that understand their company’s contracts, policies, regulatory obligations and risk thresholds without routing every question through expensive professionals. Law begins shifting from a department you consult into an environment you operate inside. And that creates a remarkable inversion. A legal department can shrink dramatically in headcount while gaining unprecedented influence over corporate behavior. A small group establishes the rules. Machines apply those rules across thousands of decisions. Human experts intervene when something exceeds the organization’s permitted boundaries. One legal professional can govern more activity than entire departments previously supervised. But the same architecture concentrates risk. Historically, legal mistakes were distributed across different people and transactions. A shared AI system can apply the same mistaken interpretation to thousands of contracts before anyone notices. The cost of legal reasoning collapses while the potential consequences of a single error multiply. That creates a new scarcity: accountable judgment. Someone still has to determine which risks the institution accepts, who can authorize exceptions and who bears responsibility when machine-generated decisions fail. And this extends far beyond law. Once sophisticated reasoning becomes embedded directly into business operations, organizations begin questioning why so much of their structure was built around waiting for experts. Accounting, consulting, finance, compliance and management all face versions of the same transformation. Institutions can become more tightly governed while becoming less dependent on the professionals who historically performed that governance. Authority concentrates while routine human intervention declines. And that brings us to the larger question. Who gets to write the rules that millions of machine-mediated decisions will obey? Whoever controls that layer gains leverage over something far more consequential than professional services. They gain influence over the operating logic of institutions themselves.

⚡️The machine economy is arriving before civilization has built a law of machine agency. That is what connects the banks and Anthropic. AI is making cognition cheap. Agents are making action cheap. Once both become abundant, the scarce resource becomes trustworthy agency. For the entire history of computing, systems could usually collapse an event into a reasonably simple causal chain: a person logged in, software executed their instruction, something happened. That chain is about to explode into thousands of intermediaries. Human intent → primary agent → planning loop → external model → retrieved data → sub-agent → browser → API → another agent → financial institution → action. At the end sits a real consequence. Money moved. Code deployed. Credentials changed. A contract was signed. A system was disabled. Yet no single node necessarily contains the complete intention that produced the outcome. That creates the deepest vulnerability: Capability is becoming distributed faster than responsibility can be assigned. Cybercrime becomes far more dangerous under that condition because an attacker no longer needs to personally execute the attack. The valuable skill becomes creating a process that searches, adapts and acts repeatedly. One human becomes the seed of an expanding machine labor force. Meanwhile legitimate agents create their own security fog. A bank sees activity coming from valid credentials. A corporation sees an authenticated process. Every individual action may look permissible while the composite behavior has wandered far beyond the original mandate. That is why these two signals fit together so cleanly. The first great crisis of agentic AI will probably be a crisis of authorship. Who actually caused the action? Who intended it? Who had authority? Where did the authority mutate? Who carries the loss? Those questions sound legal. They are about to become core computer architecture. And the response will reshape the internet. High-value systems will progressively stop trusting possession of credentials. Machines will have to establish identity, provenance, delegated authority and current permission before consequential actions are accepted. The open internet begins acquiring borders for machines. Banks become border checkpoints for financial agents. Enterprises create checkpoints around corporate systems. Governments create them around public infrastructure. Agent passports, scopes, revocation, attestation and audit trails become ordinary plumbing. Which leads to the deeper power shift: The most important infrastructure of the agentic age may become the machinery that decides which intelligences are allowed to act. Intelligence becomes abundant. Permission becomes scarce. And scarce things become power.