Brownstone Research Draws Parallel Between Decentralized AI and Bitcoin's 2014 Regulatory Battle

Brownstone Research analyst Ben Lilly contends that open-source artificial intelligence is retracing the regulatory path Bitcoin followed a decade ago, and that decentralized AI projects — which he terms 'DeAI' — could represent an investment opportunity comparable to buying Bitcoin in 2014. The argument appears in the latest edition of Lilly's newsletter, Chain of Thought.

Lilly anchors his case in testimony Anthropic CEO Dario Amodei delivered to Congress in July 2023. Amodei described open-source AI as broadly beneficial to science but cautioned that scaling open models was heading 'down a very dangerous path,' while characterizing risks from models released to date as 'relatively limited.' Lilly interprets the framing as an implicit argument that closed, commercially licensed models — such as those sold by Anthropic — are the safer regulatory choice, and that policy pressure will follow to restrict open alternatives.

The piece draws a direct line to Bitcoin's early critics. Lilly references Rep. Jared Polis purchasing the first Bitcoin on Capitol Hill in 2014 alongside Sen. Joe Manchin's call to ban what Manchin described as a 'dangerous currency.' He also invokes the 2023 allegations that U.S. regulators sought to sever crypto from the banking system — an effort critics labeled 'Operation Choke Point 2.0.' The crypto industry ultimately survived that pressure, Lilly notes, with Washington now advancing clearer frameworks through the passed GENIUS Act and the pending CLARITY Act.

Lilly identifies recent actions as evidence that similar walls are rising around open-source AI. He cites a U.S. export ban on Anthropic's latest model release, which he argues will push the company toward permissioned access requiring identity verification before a user can run a model. He also points to OpenAI's decision to limit its GPT-5.6 rollout to trusted partners. 'It's for your protection, you see,' Lilly writes. 'It always is.'

A national-security episode underpins the fear driving those moves, according to the newsletter. Lilly references NSA chief Joshua Rudd, as relayed by Sen. Mark Warner, describing how Anthropic's 'Mythos' model penetrated 'almost all of our classified system, not in weeks, but in hours.'

Despite the regulatory momentum, Lilly argues open-source models are narrowing the gap with frontier closed systems. He notes that the recently released GLM-5.2 scored on par with Anthropic's Sonnet 4.6 from February, placing open models roughly three to four months behind the frontier. He forecasts an open-source rival to both Mythos and GPT-5.6 arriving by fall 2026.

The structural argument Lilly advances centers on decentralized training over peer-to-peer networks, drawing an analogy to how Bitcoin and Ethereum swap compute for network security — except here compute is exchanged for model training. He reports that distributed training capacity has expanded from under one billion parameters to 100 billion over the past two years.

Three early-stage projects are named as examples. Dark Bloom enables low-cost private inference on idle Mac hardware. c0mpute operates as a decentralized inference network. Pluralis trains AI models across distributed consumer GPUs. Lilly anticipates additional projects will launch tokens and reward users for contributing compute resources.

The newsletter concludes with the prediction that government attempts to ban open models will ultimately fail, just as Bitcoin bans did not materialize into lasting restrictions. Lilly frames investing in decentralized AI now as analogous to acquiring Bitcoin in 2014, 'back when it was still dangerous.'