A recent academic study indicates that large language models (LLMs) now have the ability to “de-anonymize” internet users in large-scale scenarios. By analyzing publicly posted content, the models can potentially infer the real identities behind anonymous accounts. This discovery has not only raised concerns worldwide but also sparked discussions within the crypto community about “whether it is possible to uncover Satoshi Nakamoto’s true identity.”
Study Reveals: LLMs Make De-anonymization of Personal Data Easier
Titled “Large-Scale Online De-anonymization Using LLMs,” the research points out that LLMs can extract identity clues from unstructured text and perform semantic searches and comparisons within vast databases, enabling highly automated de-anonymization attacks.
The research team designed a four-stage process: Extract, Search, Reason, and Calibrate, simulating how an attacker could reconstruct personal features from publicly posted content and match them to real identities.
Overview of Large-Scale De-anonymization Framework
In experiments, researchers cross-matched Hacker News accounts with LinkedIn profiles, achieving about 45% recall at 99% precision; even with Reddit accounts, after time gaps and content filtering, the model could still identify a certain proportion of users under high-precision conditions.
Author Simon Lermen believes that LLMs do not create new identification capabilities but significantly reduce the manual effort previously required, enabling scalable de-anonymization attacks.
“Fake Name Protection” Failing? AI Challenges Online Anonymity
In the past, pseudonymity on the internet was used as a protective measure not because identities couldn’t be recognized, but because the cost of identification was too high. Lermen points out that LLMs change this: “Models can process tens of thousands of data points quickly, automating human investigative processes.”
He emphasizes that this does not mean all anonymous accounts will be immediately exposed, but rather that “as long as enough textual clues are left behind,” models may have the opportunity to reconstruct identity profiles. In other words, in the future, text could become a target for micro-data mining—signals like interests, background, or language habits could serve as identifiers even without names or account links.
Privacy Concerns in the Crypto World: Will On-Chain Transparency Become a Monitoring Tool?
This study quickly sparked discussions within the crypto community. Mert Mumtaz, co-founder of Helius Labs, believes that blockchain fundamentally relies on pseudonymous identities, and since all transaction records are permanently public, AI linking on-chain addresses to real identities could enable the creation of long-term financial activity profiles.
He worries that blockchain, originally seen as a decentralized financial infrastructure, might become a highly transparent monitoring tool in this context.
(Bitcoin’s Public Receipts No Longer Naked! How Silent Payments Achieve Convenience and Privacy)
Will Satoshi Nakamoto Be Identified by AI? Stylistic Analysis as a New Variable
Meanwhile, Nic Carter, partner at Castle Island Ventures, raised another question: if LLMs can perform advanced stylometry, could they compare Satoshi Nakamoto’s past emails, forum posts, and whitepapers to infer their true identity?
He suggests that, in theory, if comparable publicly available writing samples exist, models might perform probabilistic matching; however, this remains a statistical inference rather than a definitive proof tool. If the creator changes their writing style or has never published under their real name, identification becomes fundamentally difficult.
(Epstein Files Reveal Early Power Networks in Bitcoin; Could This Sex Offender Be Satoshi Nakamoto?)
AI’s Impact on Privacy: Encryption and Anonymity Techniques Still Need Upgrading
Lermen concludes by emphasizing that the goal is not to cause panic but to highlight the need for updates to traditional encryption and anonymity mechanisms. Previously, concerns focused mainly on structured data; now, unstructured text can also be identified. Privacy is no longer just a technical issue but involves platform policies, data disclosure habits, and social norms.
Against the backdrop of rapid AI advancements, how user privacy is redesigned and protected has become a key challenge for companies.
This article, “LLMs May Break Internet Anonymity and Privacy: Can AI Find Out Who Satoshi Nakamoto Is?” originally appeared on Chain News ABMedia.
Related Articles
Exaggeration » Korean Tax Agency Press Release Reveals Mnemonic, 6.4 Billion Won in Seized Assets Emptied Overnight
User wallet information exposed! On-chain detective ZachXBT reveals Axiom employee suspected of insider trading
XRP Ledger Dev Raises Alert on Fake 'Passes' Scam Targeting Wallets - U.Today
Bitcoin Extortion Plot Turns Violent as Fake Mailman Forces Way Into Home
The US authorities seize over $61 million USD in USDT related to "pig slaughter" scams
Six Polymarket accounts suspected of insider trading, betting on US sanctions against Iran, profit of $1 million