The origin story of one of the largest AI policy war chests in the world starts with a dog coin, a closet in Canada, and a 78-digit number.

In 2021, the creators of Shiba Inu sent a massive chunk of SHIB tokens to Vitalik Buterin's wallet without asking. The idea was simple. Put "Vitalik owns half our supply" in the marketing materials and ride the association to become the next Dogecoin. The tokens quickly ballooned in value, reaching a book value of over $1 billion.

Buterin wanted out. In a post on X on Friday, he described scrambling to liquidate before the bubble popped, including calling his stepmother in Canada and asking her to go into his closet, read out a 78-digit number, and add it to another 78-digit number transcribed from a paper in his backpack.

He sold what he could for ETH and donated $50 million to GiveWell. But he was still sitting on a mountain of SHIB.

He split the remainder in half. One half went to CryptoRelief, which used part to fund medical infrastructure in India and part to support Balvi, Buterin's own research initiative.

The other half went to the Future of Life Institute, an organization focused on existential risks from AI, biotech, and nuclear weapons. FLI had presented him with a roadmap covering all major risk categories, as well as "pro-peace and pro-epistemics initiatives."

Buterin expected FLI would cash out $10 to $25 million, given how thin SHIB's liquidity was. Instead, they managed to liquidate roughly $500 million. CryptoRelief pulled off a similar exit from its half, he said.

A meme coin that nobody took seriously had just created a billion-dollar philanthropy event, and half of it went to an organization that would soon pivot its entire strategy.

That pivot is the reason Buterin posted on Friday. He said FLI underwent "an internal pivot by which they started focusing on cultural and political action as a primary method, quite different from the original approach."

FLI's justification, according to Buterin, is that AGI is advancing rapidly, and the organization needs to move aggressively to counter the lobbying budgets of large AI companies.

He pointed to FLI's biosafety approach as an example. The organization's primary strategy has been to embed guardrails into AI models and bio-synthesis devices so they refuse to create dangerous outputs.

Buterin called this "very fragile," noting that jailbreaks, fine-tuning, and other workarounds make such restrictions easy to circumvent. He warned that the logical endpoint of that approach leads to "let's ban open-source AI" and then "let's support one good-guy AI company to establish global dominance and don't let anyone else get to the same level."

He also flagged a structural problem with regulation-first strategies. When governments restrict dangerous technology, national security organizations inevitably get exempted, and those same organizations are often a source of the risk themselves. He cited government lab leak programs as an example.

However, Buterin said he's been "heartened" by some of FLI's recent work, specifically a "pro-human AI declaration" that he said, "unites conservatives, progressives and libertarians, America, Europe and China." He also noted FLI has been researching ways to avoid the concentration of power from AI.

But the core message was clear. A donation Buterin never planned, from tokens he never wanted, funded an organization that pivoted away from the approach he believed in, and is now deploying hundreds of millions of dollars in ways that make him uncomfortable. He shared his concerns with FLI on "several occasions" before going public.

FLI did not immediately reply to CoinDesk's request for comments on the donated amount and AI safety concerns.