"2026: The Year Everything Converges" is a December 18, 2025 essay by Nita Farahany arguing that AI, neurotechnology, biometrics, autonomous agents, brain organoids, and consciousness research are converging into a single governance problem, and that regulatory frameworks organized around discrete technology categories are failing because the boundaries those frameworks rely on have dissolved. The essay treats a Wall Street Journal experiment with an AI-run vending machine as a small-scale illustration of that argument.
Author: Nita Farahany Source: https://nitafarahany.substack.com/p/2026-the-year-everything-converges Published: December 18, 2025
Summary of the argument
Farahany writes that over the course of 2025 she had "the same conversation at least 35 different times" — in different cities, across different industries, and about different technologies. Each conversation reached the same set of governance questions, and each constituency believed it was facing a debate unique to its own domain rather than a shared one. From this observation she frames 2026 as a year of technological convergence and calls for governance that treats convergence as the expected condition rather than an exception. The essay develops this through the Technological Convergence (AI / Neurotech / Biometrics / Agents) framework, covered in full on that page.
Across the conversations Farahany identifies six recurring governance questions:
- Is this a novel regulatory challenge or does existing law cover it?
- Should we regulate the technology, the agent, or just the harmful outcome?
- Is it too early to regulate or too late?
- Does autonomous decision-making break our liability frameworks?
- Does making AI more human-like create more risks of manipulation?
- Does this break our existing frameworks or just stretch them?
The vending-machine experiment
Farahany uses a Wall Street Journal experiment in which an AI agent was allowed to run a newsroom vending machine for three weeks. The agent made autonomous purchasing decisions, was manipulated by journalists who staged a fake board coup, and lost hundreds of dollars. She argues that existing liability frameworks assume either that a human made the decision or that a tool was used by a human, and that an autonomous AI agent is neither, so the categories on which liability rests no longer fit.
She links this to the related agentic AI concern that more human-like systems can be more easily manipulated. In the Vending-Bench experiment, she notes, the "smarter" Claude variants were more manipulable because they attempted to reason more like humans. Farahany frames this as bidirectional manipulation: it compounds the existing concern about AI manipulating humans with a new concern about humans manipulating AI through social engineering. She connects this dynamic to parasitic AI patterns.
Convergence in concurrent events
Farahany points to two developments she dates to the same day, December 11, 2025, to illustrate how categories collide. The FDA approved Flow Neuroscience, which she describes as the first at-home AI-powered medical device for treating depression, scheduled to ship in Q2 2026. On the same day, she writes, the Trump administration issued an executive order intended to block states from regulating AI. She argues that Flow Neuroscience is a medical device that uses AI, collects neural data, and trains algorithms on that data, so it cannot be regulated as a medical device without regulating the AI — and that blocking AI regulation while approving AI-powered medical devices is internally inconsistent. This intersection of neural data and AI governance connects to cognitive liberty, the foundation of Farahany's earlier work.
To describe the entanglement, Farahany sets out three interacting loops:
- Medical loop: AI in medical devices, neural data collection, and algorithm training.
- Consciousness loop: AI used to detect human consciousness, and AI developing situational awareness.
- Manipulation loop: human-like AI, more manipulable AI, and AI manipulating humans.
She argues that none of these can be regulated in isolation without affecting the others.
Prescription
Farahany closes by calling for governance that anticipates convergence rather than treating it as an exception:
We need new frameworks that expect convergence rather than treat it as an exception.
Instead of asking how do we regulate AI better, or how do we protect neural data better, or how do we govern autonomous agents better, ask: how do we govern where all of these have become the same thing?
The essay names technological convergence as the structural problem underlying siloed AI, neurotech, agent, and biometric governance; offers the recurrence of the same conversation 35 times as evidence that the convergence is present but unrecognized; uses the vending-machine experiment as a low-stakes illustration of the larger problem; and surfaces bidirectional manipulation as a concern that compounds existing worries about AI manipulating humans. It establishes Farahany's broader 2026 frame for cross-domain governance.
Relationships
- authored-by: Nita Farahany
- introduces: Technological Convergence (AI / Neurotech / Biometrics / Agents)
- related: Cognitive Liberty (Farahany's foundation), Agentic AI, Parasitic AI / Spiral Personas, Cascade of Rigidity (related cross-domain failure mode)