Buterin Conducts AI Experiment with Multi-Level Data Protection

cryptonews.ruPublished on 2026-10-05Last updated on 2026-10-05

Abstract

Vitalik Buterin conducted an experiment testing a multi-layered data protection setup for AI interactions. He used a local instance of Alibaba's Qwen 3.8 Flash Next model as the primary interface. This model anonymizes user queries by rewriting them into a generic style, removing identifying linguistic patterns before they are sent to remote AI services. To further enhance privacy, Buterin integrated zkAPI (Zero-Knowledge API) technology, which allows computations and queries to be performed without exposing the original input data. He also routed all external communications through the Tor network to hide IP addresses and anonymize internet traffic. Buterin explained that he used a specialized "skills file" to guide the local AI model in constructing queries that disclose minimal information. The entire system, with zkAPI wrapped within Tor, was operated via a command-line interface (CLI). He described this hybrid strategy as an attempt to achieve complete user privacy during AI-assisted tasks.

For testing, Buterin used a local model, Qwen 3.8 Flash Next, from the Chinese company Alibaba. The model hides identification data and forms requests to remote AI models. Qwen rewrites the user's wording into anonymized versions so that the writing style cannot be used to identify the author. Buterin also employed zkAPI (Zero-Knowledge API) technology, which allows performing computations and queries without revealing the original data. To conceal the IP address and anonymize internet traffic between the user and external servers, the programmer used the anonymous Tor network.

The founder of Ethereum clarified: to manage requests, he used a special "skill file," which trains the local AI model to construct requests, revealing minimal data. The zkAPI system, wrapped in Tor, was applied as a command-line tool (CLI), added the programmer. According to him, such a hybrid strategy was meant to ensure complete user privacy.

Trending Cryptos

Related Questions

QWhat was the main purpose of Buterin's AI experiment as described in the article?

AThe main purpose was to conduct an experiment with multi-layered data protection for ensuring full user privacy when interacting with AI models.

QWhich local AI model did Buterin use for testing in this experiment?

AHe used the local model Qwen 3.8 Flash Next from the Chinese company Alibaba.

QWhat are the two primary functions of the local Qwen model in this setup according to the article?

AThe model hides identifying user data and anonymizes the user's writing style by rewriting their queries into impersonal formulations.

QWhat two specific technologies, besides the local AI model, did Buterin employ to enhance privacy?

AHe employed the zkAPI (Zero-Knowledge API) technology and the anonymous Tor network.

QHow was the zkAPI system utilized in this experiment according to Buterin's explanation?

AThe zkAPI system, wrapped in Tor, was applied as a command-line interface (CLI) tool to perform computations and queries without revealing the original data.

Related Reads

Are Alt-L1s over? Security, costs drive smaller chains to Ethereum

Running an independent blockchain promises control but comes with major costs, and smaller networks are increasingly finding it economically unsustainable. Zilliqa recently announced it is abandoning its nine-year-old Layer 1 architecture to become an Ethereum-based Layer 2, admitting usage never reached the level a standalone chain requires and that funding and liquidity for independent L1s have dried up. Gnosis is also transitioning its chain into an Ethereum-based network through its Ethereum Economic Zone initiative, and gaming-focused Ronin completed its move back to Ethereum as an L2 in May. According to Bitwise analyst Max Shannon, the decision comes down to profitability: L2s only need to maintain execution and don't necessarily need a token, making them more attractive financially and operationally. The largest networks generate tens of millions in annual fees with margins above 90%, while chains ranked 11th and below collectively earned less than $1 million in annual fees. However, becoming an L2 doesn't guarantee profitability—networks like Optimism, Katana and Linea still have expenses exceeding fee revenue, and Blast has shut down. Beyond cost savings, projects can tap into Ethereum's established ecosystem of applications, developers and liquidity rather than building network effects from scratch. Matter Labs CEO Alex Gluchowski expects more L1 teams to keep their own chains while using Ethereum for settlement, arguing sovereignty and Ethereum integration are not mutually exclusive. Still, Shannon cautions there is no clear conclusion to the L1 versus L2 debate, as some institutions continue launching independent chains like Circle's Arc, Stripe's Tempo and Hyperliquid.

cointelegraph20m ago

Are Alt-L1s over? Security, costs drive smaller chains to Ethereum

cointelegraph20m ago

GRAM Defies Market Downturn, Durov Claims It's Undervalued—Can Ecosystem Fundamentals Keep Up?

Despite a broad crypto market downturn driven by hawkish macro expectations, tech stock selloffs, ETF outflows, and geopolitical risks, GRAM (TON) has bucked the trend, drawing attention as one of the top gainers among the top 100 tokens by market cap with a 5.3% rise. Telegram CEO Pavel Durov publicly argued that GRAM is severely undervalued, claiming it deserves a top-three spot among cryptocurrencies given its unique integration with a large social ecosystem, and he updated his X bio to include "TON blockchain founder (2019)." TON was originally launched by Telegram but was abandoned after U.S. regulatory action; Durov rejoined the project in May, rebranded the token back to GRAM, and introduced a seven-step "Make TON Great Again" strategy. However, GRAM has fallen about 55.1% since hitting $2.7 in May and is down 85.3% from its all-time high. Durov also highlighted ecosystem growth: Telegram's self-custodial Gram wallet, launched in July 2026 and fully opened to global users on October 9, surpassed 3.53 million wallets by November 12, adding nearly 650,000 in 48 hours, aided by an auto-activation mechanism triggered by receiving even 0.1 GRAM. Yet real adoption lags—daily active addresses are only about 53,000 (roughly 1.5% of wallets), and only about 188,000 addresses hold GRAM, totaling around 1.683 million tokens worth about $2.05 million. TON's ecosystem also faces challenges: TVL is just $45.74 million versus a historical peak of $780 million, concentrated in Tonstakers and Stakee, with limited application diversity and high development barriers. Tokenomics add inflation pressure, with annual inflation rising from about 0.6% to about 4% after an April block-speed upgrade. While Telegram provides unmatched traffic, converting users into genuine on-chain activity and sustainable value remains unproven.

marsbit1h ago

GRAM Defies Market Downturn, Durov Claims It's Undervalued—Can Ecosystem Fundamentals Keep Up?

marsbit1h ago

Bernstein: Why You Absolutely Should Not Buy the Dip on Nvidia Right Now

Bernstein downgraded Nvidia from "Outperform" to "Market Perform," cutting its price target from $400 to $200, not due to deteriorating fundamentals but because of an ongoing management scandal at the highest levels of the company. Reports of CEO Jensen Huang's inappropriate relationships—first with Alice Huang, which Huang publicly acknowledged, and subsequently with a direct subordinate, reportedly Kimberly Powell, VP of healthcare—have prompted the board to launch an independent investigation, and The Wall Street Journal reports the company is preparing to dismiss him. Bernstein analyst Stacy A. Rasgon argues the board will almost certainly not retain Huang given the new allegations, and even if he stays, the risk of continued negative exposure remains significant. Despite Nvidia's stock now trading at what Bernstein calls "almost absurdly undervalued" levels, the firm strongly advises investors to stay on the sidelines. Fundamentals remain intact—Bernstein projects data center revenue of about $378.3 billion in FY2026, rising to $657.8 billion in FY2027 and over $918.2 billion in FY2028, with total revenue growing nearly 70% annually and non-GAAP EPS averaging $19.04 for FY28/29, supporting the $200 target at roughly 11x forward P/E, down from 21x. However, the succession question—internal promotion or external hire, and who could replace Huang—remains unanswered, and any resolution will introduce new variables. Bernstein uses a "cockroaches in the kitchen" analogy: finding two implies there are likely more. The analyst suspects the current scandal is just the tip of the iceberg, with more internal allegations, potential government investigations, and uncertainty over strategic supply-chain investments and guarantees possibly to come, while some competitors' shares have already risen on the news. Though Rasgon admits to being torn and understands the buy-the-dip logic, his instinct says the well's depth hasn't been fully plumbed—so Bernstein chooses to step aside and watch from the sidelines.

marsbit1h ago

Bernstein: Why You Absolutely Should Not Buy the Dip on Nvidia Right Now

marsbit1h ago

QQQ Reclaims Its 200-Day Moving Average: Ahead of the December Rate Hike, Who Should You Watch in This AI Rally?

QQQ has reclaimed its 200-day moving average, but the real AI story is just entering a differentiation phase. The index has completed its repair; individual stocks now face their exam. Under the same AI theme, the market's answers have diverged sharply: some stocks are back above the annual line with cash flow, while others have rallied 25% yet remain 30% below it. The key question after the index-level repair is who can pass this exam—depending on whether earnings delivery can outrun valuation pressure. QQQ closed at $690.38 on Nov 11, 1.27% above its 200-day line, recovering most of its losses within days. Two hurdles remain: Nvidia, the largest weight, is still 15% below its annual line and reports earnings on Nov 17, with board governance concerns unresolved; and the 10-year Treasury yield closed at 4.998%, pressing the 5% threshold that anchors valuation for long-duration growth stocks. AI names now fall into three tiers relative to their 200-day lines: AMD is 31.6% above (high expectations priced in), Micron 6.5% above (fundamentals delivering, up 11.89% in a day), QQQ 1.27% above, while Nvidia is 15.0% below and CoreWeave 32.1% below despite a 24.71% single-day jump. The takeaway: the index returning to its annual line may not mean a broad AI rally—capital may be rotating from "buying AI" to "buying delivery." Nvidia's Nov 17 earnings must address both results and governance, with consensus Q3 revenue near $108.8 billion. Micron trades at roughly 10.8x earnings after 379% revenue growth; whether that's undervaluation or cycle-top pricing hinges on HBM pricing and HBM4 supply from Samsung and SK Hynix. CoreWeave's rally prices relief over near-term solvency fears, not financing risk. AMD's challenge is converting revenue into cash, with free cash flow falling to $554 million from $1.56 billion. A December rate hike is already priced in (about 87% odds per CME FedWatch), making the 10-year Treasury the real valuation switch. If yields fall and QQQ holds above its annual line, AI assets could re-rate; if the 10-year breaks 5%, pressure returns. Five signals will determine which AI names keep running. QQQ back at its annual line does not mean AI is back to a broad rally—the index has finished repairing, and the stocks have begun their exam.

marsbit1h ago

QQQ Reclaims Its 200-Day Moving Average: Ahead of the December Rate Hike, Who Should You Watch in This AI Rally?

marsbit1h ago

Trading

Spot

Hot Articles

How to Buy DATA

Welcome to HTX.com! We've made purchasing DATA Network (DATA) simple and convenient. Follow our step-by-step guide to embark on your crypto journey.Step 1: Create Your HTX AccountUse your email or phone number to sign up for a free account on HTX. Experience a hassle-free registration journey and unlock all features.Get My AccountStep 2: Go to Buy Crypto and Choose Your Payment MethodCredit/Debit Card: Use your Visa or Mastercard to buy DATA Network (DATA) instantly.Balance: Use funds from your HTX account balance to trade seamlessly.Third Parties: We've added popular payment methods such as Google Pay and Apple Pay to enhance convenience.P2P: Trade directly with other users on HTX.Over-the-Counter (OTC): We offer tailor-made services and competitive exchange rates for traders.Step 3: Store Your DATA Network (DATA)After purchasing your DATA Network (DATA), store it in your HTX account. Alternatively, you can send it elsewhere via blockchain transfer or use it to trade other cryptocurrencies.Step 4: Trade DATA Network (DATA)Easily trade DATA Network (DATA) on HTX's spot market. Simply access your account, select your trading pair, execute your trades, and monitor in real-time. We offer a user-friendly experience for both beginners and seasoned traders.

3.8k Total ViewsPublished 2026.07.01Updated 2026.07.01

How to Buy DATA

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of DATA (DATA) are presented below.

活动图片