Ethereum co-founder Vitalik Buterin has publicly confirmed that artificial intelligence successfully pierced his attempt at online anonymity, validating concerns about whether pseudonymous technical contributions can survive modern AI analysis.
According to reporting by CoinTelegraph, Co-Invest CEO Franklyn Wang identified Buterin as the author of an anonymous rewrite of Ethereum Improvement Proposal (EIP-7503) with roughly 20% confidence—about 10 times higher than any other candidate Wang’s analysis examined.
The experiment carries real implications for how blockchain communities operate. Developers from Bitcoin’s pseudonymous creator Satoshi Nakamoto to modern Ethereum contributors have relied on anonymity to separate technical merit from personal identity. If AI can reliably identify authors through reasoning patterns rather than prose style, that privacy shield deteriorates.
The Challenge and the Crack
Two weeks before Wang’s successful identification, Buterin issued a public challenge on June 22. He claimed to have published a “medium importance” document related to Ethereum sometime in the past decade under a different identity. “Find it,” he wrote, testing whether current AI tools could pierce his anonymity.
Wang’s team tackled the problem methodically. They analyzed 27 documents and ranked candidates by likelihood of authorship. The winning indicator wasn’t Buterin’s word choice or phrasing—Wang had specifically tried to obscure those by writing in Chinese, machine-translating to English using Qwen 2.5, and manually correcting the translation.
What Wang’s AI actually detected was deeper: the intellectual patterns underlying Buterin’s mathematical and technical reasoning. “The tell wasn’t his words, it was his reasoning,” Wang wrote on X after Buterin’s confirmation.
Why This Matters for Ethereum Development
Buterin later acknowledged the vulnerability in his strategy. He confirmed he had published the anonymous EIP-7503 rewrite and explained that his obfuscation covered surface-level prose only. “His AI picked up on intellectual habits and style of math and algorithm explanation, which bypassed my obfuscation strategy completely,” he noted.
The implication extends beyond one experiment. A February research paper from ETH Zurich and Anthropic found that large language models can extract identity-related information from unstructured text and match it against potential candidates, outperforming traditional deanonymization techniques at scale.
Vladimir Novakovski, CEO of Lighter, said his team previously attempted similar analysis to identify Bitcoin’s Nakamoto using GPT-4 to match writing styles in cryptography research. That effort failed to produce high-confidence results. Wang’s success against Buterin represents a significant step forward in AI-assisted author identification.
The Broader Anonymity Question
The development raises practical questions for blockchain communities. Open-source projects have long benefited from anonymous contributions—they allow technical merit to stand alone without personal reputation or politics clouding review. If reasoning patterns become as identifiable as fingerprints, that separation collapses.
Buterin’s willingness to test the boundaries himself suggests he views the finding as important information for the Ethereum ecosystem. Rather than treating anonymity as inviolable, acknowledging AI’s deanonymization capability allows developers to make informed choices about how much identity obfuscation is worth attempting.
The challenge ahead isn’t preventing AI identification altogether—that appears increasingly impractical. The real work involves deciding which technical contributions actually require anonymity, and which communities can operate transparently without sacrificing the principle of merit-based evaluation.