https://joviantan.com https://joviantan.com/images/jovian-tan/corporate-portrait-negative-space.png Dr. Jovian Tan smiling, standing with hands in pockets in a white shirt, charcoal waistcoat, red patterned tie and black-rimmed glasses, against a dark studio background https://joviantan.com/about https://joviantan.com/images/jovian-tan/jovian-tan-portrait-studio-alt.png Portrait of Dr. Jovian Joris Tan Chern Chian https://joviantan.com/images/jovian-tan/jovian-tan-portrait-suit.png Dr. Jovian Joris Tan Chern Chian smiling and buttoning a dark three-piece suit https://joviantan.com/insights/who-controls-the-exit-crypto-exchanges-gatekeepers https://joviantan.com/images/insights/who-controls-the-exit-crypto-exchanges-gatekeepers.png Who Controls the Exit? When Crypto Exchanges Become the Gatekeepers of Your Money Illustration titled Are Exchanges Now the Gatekeepers of Your Money?, with the banner Binance Blocks MEXC and the line The Binance–MEXC Move, $16B Regulatory Scrutiny, and What It Means for Crypto Users. A stream of gold coins crosses a bridge between a Binance gate and a MEXC gate and stops at a glowing red barrier marked Transfer Blocked https://joviantan.com/images/insights/exchange-exit-key-dates.png A government order, a published rule and a separate platform decision. Six cards headed Binance, MEXC and FinCEN: the key dates. 6 November 2026: FinCEN order announced under section 9714(a), concerning Russian illicit finance. 7 November 2026: MEXC responds that it is reviewing the order and operating normally. 10 November 2026: the order is published in the Federal Register, 91 FR 72420. 10 November 2026: Binance restricts transactions, declining sends and holding deposits for review. 10 December 2026, highlighted: the order takes effect for covered US financial institutions. Alleged, not adjudicated: at least US$16 billion, Treasury's estimate for May 2021 to May 2026 https://joviantan.com/images/insights/exchange-exit-access-layers.png Conceptual, not a specific transaction route. Diagram headed Where access is controlled, a simplified map of the crypto-to-fiat journey. Four gates in a row, each with who can say no. Public blockchain, which moves and settles tokens: protocol rules. Exchange or custodian, highlighted, which credits, holds or declines deposits: exchange compliance. Stablecoin issuer, which mints and redeems tokens: issuer controls. Bank or fiat gateway, which converts to bank money: banks and regulators. Below: an open protocol does not oblige any intermediary to accept a transfer https://joviantan.com/insights/the-new-financial-order https://joviantan.com/images/insights/the-new-financial-order.png The New Financial Order: Is Blockchain Quietly Rebuilding the Global Banking System? Illustration titled The New Financial Order: Is Blockchain Quietly Rebuilding the Global Banking System?, with the line From Tokenised Deposits to Stablecoins, Banks Are Moving Onto Blockchain. A classical bank building stands beside a glowing globe and city towers, with a chain of glowing blocks running across a world map https://joviantan.com/images/insights/agora-real-value-test.png Reported test values, not market-wide adoption figures. Six cards headed Project Agorá: the July 2026 real-value test, as reported by the BIS, a controlled test and not a commercial benchmark. Participants: 28 financial institutions and central banks. Scenarios: 17, with 30 transactions in selected currencies. Total value: about CHF 800,000 across all test transactions. Range: CHF 9,000 to 125,000 per transaction, or local equivalent. Speed: about 80 seconds on average from initiation to settlement. Scope: a controlled test, with the prototype not integrated with RTGS systems. Source: BIS Innovation Hub, Project Agorá project page, checked 9 November 2026 https://joviantan.com/images/insights/digital-finance-architecture.png Conceptual. Specific platforms use different designs, governance and settlement assets. Diagram headed The emerging architecture of digital finance, a conceptual model, in five layers. Users: individuals, businesses and financial institutions. Payments and settlement: domestic, cross-border and wholesale. Tokenised money, highlighted: tokenised deposits, regulated stablecoins and wholesale CBDC. Shared ledger infrastructure: shared state, programmability and interoperability. Regulation and trust: central banks, supervisors, law and compliance. Note: each form of money remains a different legal claim, whatever ledger it sits on https://joviantan.com/images/insights/three-forms-of-digital-money.png Same kind of ledger, different legal claims. Three columns headed Three forms of digital money: similar digital interfaces can represent very different legal claims. Tokenised deposit: issued by a commercial bank as a claim on that bank; main risks are bank credit, liquidity and operations; the key question is whether the bank is sound. Regulated stablecoin: issued by a private firm, with rights set by law and terms; main risks are reserves, redemption and custody; the key question is whether holders can redeem at par. Wholesale CBDC: issued by a central bank as a claim on the central bank; main risks are design, access and policy; the key question is who can access it. Below: the legal claim matters more than the ledger it sits on https://joviantan.com/images/insights/tokenised-cross-border-settlement.png What changes, and what does not. Diagram headed Cross-border payments: what tokenisation changes, simplified and illustrative, in three rows. Correspondent model: the sender's bank sends payment messages, correspondent banks keep separate ledgers with cut-off times, and the recipient's bank reconciles after the fact. Shared ledger: tokenised deposits and reserves move on a programmable platform with conditions in smart contracts, and settle atomically, so all legs complete or none do. Still required in both: compliance checks such as KYC, sanctions and AML/CFT; liquidity and FX funding; and legal finality, set by law, not by code. Below: tokenisation changes how a payment is coordinated, not the obligations behind it https://joviantan.com/insights/ueno-bank-collapse-crypto-infrastructure https://joviantan.com/images/insights/ueno-bank-collapse-crypto-infrastructure.png Ueno Bank’s Collapse: What It Really Tells Us About Crypto Infrastructure Illustration titled Ueno Bank's Collapse: What It Really Tells Us About Crypto Infrastructure. A cracked classical bank building labelled Ueno Bank faces a glowing glass vault with blockchain cubes across a bridge, with a city skyline and a rising candlestick chart behind https://joviantan.com/images/insights/ueno-deposit-guarantee.png The statutory cap at the time of the central bank's announcement. Diagram headed Paraguay's deposit guarantee: what the cap means, not drawn to scale. A bar is split at the cap: covered eligible deposits, paid through the guarantee scheme, and the balance above the cap, which depends on the resolution process. The cap is 228.3 million guaraníes, 75 monthly minimum wages, per depositor, per institution, not per account, and is not a promise of instant access or of full recovery above the limit. Source: Banco Central del Paraguay, 3 November 2026 https://joviantan.com/images/insights/ueno-key-facts.png What the regulator and the rating agency have stated. Six cards headed Ueno Bank: key facts. Decision: on 2 November 2026 the central bank's board decided to resolve the bank. Trigger: solvency ratio below 50% of the regulatory minimum after adjustments to its financial statements. Rating: FIX SCR downgrade from AA-py to Dpy on 3 November 2026. Concerns: related-party operations, asset quality and documentation. Guarantee cap: 228.3 million guaraníes per depositor, per institution. Status: resolution under way, recovery above the cap not yet known. Sources: Banco Central del Paraguay, 3 and 5 November 2026; FIX SCR, 3 November 2026 https://joviantan.com/images/insights/ueno-risk-comparison.png A qualitative framework, not a ranking. Three columns headed Three models, different risk profiles. Deposit-taking bank: the bank owes deposits and deploys funds on its balance sheet; prudential rules on capital, liquidity and supervision; a deposit guarantee for eligible deposits, within limits. Centralised crypto platform: the platform controls keys and may owe customers assets; risk of lending, pledging or rehypothecating them; protection varies, usually with no deposit insurance. Non-custodial: the user controls keys, with no intermediary credit exposure; transactions can be verified on-chain; the user bears key-management risk such as loss, phishing and contract bugs https://joviantan.com/images/insights/ueno-customer-asset-flow.png Where control and exposure sit in each model. Diagram headed Customer asset flow: bank versus crypto, in three rows. Deposit-taking bank: customer deposits, the bank holds and lends them, loans and returns carrying credit and liquidity risk. Centralised crypto: the customer transfers assets, the platform holds the keys, the assets may be traded or lent. Non-custodial: the user holds the keys, a smart contract executes, settlement is on-chain and verifiable. Below: who controls the asset, and who owes the customer? https://joviantan.com/insights/ai-learning-how-to-hack-cybersecurity-defence https://joviantan.com/images/insights/ai-learning-how-to-hack-cybersecurity-defence.png AI Is Learning How to Hack. Are We Ready? Illustration titled AI Is Learning How to Hack. Are We Ready? How AI is reshaping cybercrime and the future of cybersecurity. A glowing digital brain with a padlock chip, a globe and a row of security shields above a hooded figure facing screens of threat icons https://joviantan.com/images/insights/ai-hack-threat-landscape.png AI is both an attacker's tool and a new target. Six cards headed Two directions of AI cyber risk: AI used to attack (phishing, impersonation, synthetic media, faster research); AI systems attacked (prompt injection, data leakage, poisoned retrieval, tool abuse); human trust exploited (urgency, authority, familiarity and process bypass); infrastructure exposed (credentials, APIs, cloud services, endpoints and data stores); AI used to defend (alert triage, anomaly detection, investigation and response); and the core principle: verify identity, restrict authority and monitor every critical action https://joviantan.com/images/insights/ai-hack-attack-surface.png The model is only one part of the attack surface. Six numbered cards headed Where an AI application can be attacked: input and prompts (user prompts, webpages, emails and uploaded files); data and retrieval (knowledge bases, vector stores, permissions and provenance); model and context (system instructions, context windows, model supply chain); tools and APIs (email, databases, code execution and payment integrations); output and action (validation, policy checks, approvals and audit records); operations (logging, monitoring, incident response and rollback). Below: the model is only one component; the whole workflow needs security controls https://joviantan.com/images/insights/ai-hack-dbir-initial-access.png General breach statistics, not attributed to AI. Bar chart headed Known initial access vectors in the 2025 DBIR, on a common scale: credential abuse 22%, exploitation of vulnerabilities 20%, phishing 16%. Callout: not AI-specific; general breach entry vectors, not attacks caused by AI; ransomware was present in 44% of breaches, against 32% a year earlier. Source: Verizon, 2025 Data Breach Investigations Report https://joviantan.com/images/insights/ai-hack-eight-layers.png Security is a system of controls, not a single product. Eight numbered cards headed The eight-layer AI cybersecurity defence: identity (phishing-resistant MFA and verified channels); least privilege (narrow scopes, short-lived credentials, fast revocation); approval gates (dual control for payments and high-impact actions); data protection (classification, encryption, access-aware retrieval); secure AI design (untrusted-input handling and output validation); monitoring (logs, anomaly detection and response ownership); testing (red teaming, prompt-injection and permission tests); recovery (backups, containment and agent shutdown procedures) https://joviantan.com/images/insights/ai-hack-incident-response.png Decide who can act, and how, before an incident. Diagram headed A practical incident response sequence, in six steps: detect abnormal access, tool calls, exports or payments; contain by suspending the agent, account, integration or endpoint; revoke credentials, tokens, sessions and API keys; investigate by preserving logs, establishing scope and tracing actions; recover trusted systems and verify integrity; improve by fixing the root cause, updating controls and rehearsing again. Below: rehearse the sequence before you need it https://joviantan.com/insights/ai-deepfakes-financial-trust-infrastructure https://joviantan.com/images/insights/ai-deepfakes-financial-trust-infrastructure.png Can You Still Trust Your Eyes? Illustration titled Can You Still Trust Your Eyes? How AI deepfakes are breaking the world's trust infrastructure. A man's face is split down the middle, one half real skin and the other a glowing digital wireframe, beside a laptop showing a bank icon and stacks of coins https://joviantan.com/images/insights/deepfake-attack-chain.png How synthetic media becomes a financial loss. Diagram headed The deepfake attack chain, in five steps: synthesise voice, face or identity material; establish trust by impersonating a trusted person; create urgency through authority, secrecy and time pressure; bypass checks by pushing a person past normal controls; financial loss through a transfer, account takeover or data loss. Below: a convincing appearance is not proof of identity or authority https://joviantan.com/images/insights/deepfake-fbi-ai-complaints.png Four separate measures, not one scale. US complaint data only. Four separate figure cards headed AI-related cybercrime reported to the FBI in 2025: 22,364 complaints reporting AI-related information; US$893 million in adjusted losses in those complaints; more than US$632 million in investment losses with a reported AI nexus; and about US$13 million in AI-involved employment scam losses. Note: different units and categories that may overlap, not to be added together. Source: FBI Internet Crime Complaint Center, 2025 Internet Crime Report https://joviantan.com/images/insights/deepfake-payment-workflow.png Each step checks something the others do not. Diagram headed A safer high-value payment workflow, in five steps: receive the request, never relying on video or email alone; call back on a number already on file; check the payee, confirming the beneficiary and any changes; dual approval by two separate approvers; execute and log, applying limits and keeping audit evidence. Below: if any check fails, stop, escalate and preserve evidence https://joviantan.com/images/insights/deepfake-trust-layers.png Five independent layers of financial trust. Diagram headed Five layers of financial trust, each with a core question and an example control: identity (who is this? verified identity and credential checks), channel (is this the approved channel? pre-registered contact and secure portal), authority (can this person approve it? role-based permissions and dual control), transaction (what exactly is being approved? signed transaction details and beneficiary checks) and evidence (can the decision be reconstructed? tamper-evident logs and incident records) https://joviantan.com/insights/financial-system-24-7-bank-run-liquidity-risk https://joviantan.com/images/insights/financial-system-24-7-bank-run-liquidity-risk.png The Financial System Is Going 24/7. But Is It Ready for a Bank Run? Illustration titled The Financial System Is Going 24/7. But Is It Ready for a Bank Run? with the line As finance goes always-on, can liquidity, trust and regulation keep up? A glowing globe circled by a clock-like arrow stands beside a columned bank building, above glass panels holding coins, a house, gold bars and a certificate https://joviantan.com/images/insights/bank-run-transaction-chain.png A 24/7 market is a chain of dependencies. Diagram headed The 24/7 financial transaction chain: five linked steps, trade (orders and pricing), clear (obligations matched), settle (cash and asset change hands), custody (ownership recorded) and redeem (converted into usable money). Below: continuous access is not the same as continuous capacity https://joviantan.com/images/insights/bank-run-svb-withdrawals.png Nearly a quarter of SVB's deposits were requested in about eight hours. Bar chart headed Silicon Valley Bank: the speed of a digital run. Total deposits of about $166 billion beside withdrawal requests of about $42 billion on 9 March 2023, made in about eight hours. Callout: nearly a quarter of deposits, about 25%. Source: California DFPI, 8 May 2023 https://joviantan.com/images/insights/bank-run-liquidity-mismatch.png The token, the asset and the cash move at different speeds. Diagram headed The liquidity mismatch in tokenised assets, in three layers. Token layer: moves around the clock, with potentially continuous transfer and trading. Asset layer: moves more slowly, with slower valuation, limited buyers and longer liquidation periods. Cash layer: runs on banking time, with banking hours, payment cut-offs, liquidity limits and redemption terms. Below: transferability does not guarantee immediate liquidity https://joviantan.com/images/insights/bank-run-liquidation-spiral.png How automated liquidations can feed on themselves. Diagram headed How a liquidation spiral can form, a possible feedback loop in leveraged DeFi markets: prices fall, collateral weakens, smart contracts trigger liquidations, collateral is sold into a falling or thin market, and prices fall further, which can trigger the next round. Below: severity depends on collateral, oracles, market depth and protocol design https://joviantan.com/images/insights/bank-run-resilience-layers.png Seven layers a 24/7 system has to secure. Seven cards headed Seven layers of 24/7 financial resilience: asset quality (transparent, high-quality assets and realistic valuation), liquidity (stress-tested buffers and credible redemption), custody (segregation, controls, recovery and oversight), settlement (legal finality and safe settlement assets), technology (cyber resilience, redundancy and monitoring), governance (clear accountability and tested contingency powers) and regulation (cross-border cooperation and data sharing) https://joviantan.com/insights/when-ai-starts-moving-money-who-is-accountable https://joviantan.com/images/insights/when-ai-starts-moving-money-who-is-accountable.png When AI Starts Moving Money: Who Is Accountable? Illustration titled When AI Starts Moving Money: Who Is Accountable? A humanoid AI figure holds a glowing digital wallet beside a globe, Bitcoin, Tether and dollar coins, a payment-executing screen, a security shield and a bank icon https://joviantan.com/images/insights/ai-payments-boe-autonomy.png Deployment is widespread; broad autonomy is rare. Survey respondents only. Bar chart headed AI agent autonomy in UK financial services: of survey respondents with AI agents deployed, 80% use them for decision support, 46% with constrained execution, 22% with conditional autonomy and 2% with broad autonomy. A separate callout: even including planned deployment, no more than 9% expect agents in selected high-impact functions such as initiating or executing payments. Source: Bank of England and FCA, October 2026 https://joviantan.com/images/insights/ai-payments-three-layers.png The three layers of an agentic payment. Diagram headed The three layers of an agentic payment. Intent and planning: the agent interprets the goal and prepares a structured payment request. Authorisation: independent checks on identity, scope, payee, amount, limits and approvals. Settlement: a bank, card network, provider or blockchain executes and records it. Below: the agent proposes, controls authorise, infrastructure settles https://joviantan.com/images/insights/ai-payments-risk-map.png Where the risks of agentic payments concentrate. Six cards headed Six risk categories in AI-driven money: operational risk (faulty decisions, wrong payees or unintended transactions), security risk (prompt injection, account takeover, key or data compromise), financial risk (unauthorised payments, losses or liquidity shocks), legal and regulatory risk (unclear liability, consumer protection and compliance duties), governance risk (excessive permissions, weak oversight or broken audit trails) and reputational risk (loss of customer trust, disputes and brand damage). Below: prevent, limit, monitor, pause and assign accountability https://joviantan.com/images/insights/ai-payments-secure-architecture.png A safer architecture for autonomous payments. Diagram headed A safer architecture for AI-driven payments, in three layers. Application: the AI agent interprets intent, analyses options and prepares payment instructions. Control: identity and policy checks verify the principal, permissions, limits, payee, risk and approvals. Execution: payment and settlement through banks, card networks, providers or on-chain systems. Required controls: a named authorising principal, transaction limits and scope, monitoring and revocation, and an audit trail with accountability https://joviantan.com/insights/stablecoins-new-financial-infrastructure https://joviantan.com/images/insights/stablecoins-new-financial-infrastructure-v2.png Stablecoins: The New Financial Infrastructure? Illustration titled Stablecoins: The New Financial Infrastructure? How USDT, USDC and Tokenised Money Are Reshaping Global Payments. USDT and USDC coins stand before a glowing globe and a bank building, beside panels labelled global payments, cross-border settlement, RWA tokenisation and programmable money https://joviantan.com/images/insights/stablecoin-trust-stack.png The stablecoin trust stack: each layer has to hold. Diagram headed The stablecoin trust stack, showing six layers the payment promise rests on: issuer governance, reserve quality and custody, legal claim and redemption, banking access, blockchain and smart contracts, and wallet or intermediary controls. Below: a weakness in one layer can undermine the whole payment promise https://joviantan.com/images/insights/stablecoin-market-share.png Approximate market share, rounded. Estimates vary by source and date. Bar chart headed Two issuers dominate the market: USDT and USDC combined hold about 90% of the stablecoin market and other issuers about 10%. Source: IMF Departmental Paper No. 2026/021, October 2026, rounded https://joviantan.com/images/insights/stablecoin-lifecycle.png A simplified lifecycle. Real issuance and redemption depend on the issuer's terms. Diagram headed How a reserve-backed stablecoin moves, with five linked steps: fiat funding, where the customer funds the issuer; mint, where tokens are created; on-chain transfer from wallet to wallet; redeem, where an eligible holder submits tokens; and fiat payout by the issuer or a partner. Minting and redemption depend on issuer terms, eligibility checks and banking access https://joviantan.com/images/insights/stablecoin-reserve-model.png The reserve model, simplified. Composition and legal claims vary by issuer. Diagram headed The reserve model: what supports the token? The issuer creates and redeems tokens; the reserve pool holds cash, deposits and T-bills; custody and banks provide safekeeping and cash rails; the holder owns contractual rights. Trust requires reserve quality, legal rights and credible redemption. Actual reserve composition and legal claims vary by issuer https://joviantan.com/images/insights/stablecoin-payment-rails.png Stablecoin and correspondent-banking rails compared. Comparison headed Two payment rails, different trade-offs. Stablecoin rail: often 24/7, on-chain confirmation, smart-contract logic, off-ramp or merchant access. Correspondent banking: bank hours and cut-offs, multi-bank process, legacy integration, local bank rails. End-to-end cost, compliance and conversion matter more than speed alone https://joviantan.com/images/insights/stablecoin-risk-map.png Where stablecoin risks concentrate. Map headed Where stablecoin risks concentrate. Issuer and asset layer: reserve and liquidity, redemption access, bank or custodian. Technology and market layer: smart contract or chain, freeze or governance, depeg or market. Controls should span issuer, intermediary, wallet and network https://joviantan.com/images/insights/stablecoin-future-money.png Three forms of digital money, likely to coexist. Diagram headed The future is likely to be hybrid. Stablecoins (a private issuer claim), tokenised deposits (a commercial bank liability) and central bank money (a central bank liability) all connect to interoperable financial rails for payments, settlement and treasury https://joviantan.com/insights/tokenisation-liquidity-illusion-rwa-secondary-markets https://joviantan.com/images/insights/tokenisation-liquidity-illusion-rwa-secondary-markets.png The Liquidity Illusion Illustration titled The Liquidity Illusion: Why Tokenised Assets Don't Automatically Become Liquid? A glass block labelled RWA token holds a modern apartment building, above a broken bridge with a question mark between two cliffs labelled Tokenisation and Liquidity https://joviantan.com/images/insights/liquidity-test.png The RWA liquidity test. An illustrative framework, not a scoring model. Diagram headed The RWA liquidity test, an illustrative framework with five linked stages an investor must pass to complete an exit. Legal rights: can the token legally transfer? Buyer access: who is eligible to buy? Market depth: are there real bids available? Execution: at what price and quantity? Settlement: can the exit actually complete? Below: transferability is not liquidity, and a weakness at any stage can undermine the ability to exit https://joviantan.com/images/insights/liquidity-study-sample.png The Swinkels (2023) sample: historical, small and concentrated in one city. Bar chart headed Where the 58 studied properties were located: 52 in Detroit and 6 in other US locations. Source: Swinkels (2023), 58 US residential rental properties tokenised from October 2019 to February 2021 https://joviantan.com/images/insights/liquidity-research-evidence.png Selected findings from two separate studies. The figures are not directly comparable. Three figures headed Selected findings from two separate studies. Swinkels (2023): a median of 254 owners per property, and about one year on average between ownership changes. BIS Working Paper 1311: a 35% cumulative rise in trading over two days after disasters, specific to the studied US platform data and buyback features. The figures are not directly comparable https://joviantan.com/images/insights/liquidity-exit-routes.png Three routes, with different counterparties, conditions and risks. Diagram headed How an investor actually exits a tokenised asset. The issuer creates tokens and investor A subscribes in the primary route. Investor A can sell to investor B, an eligible secondary buyer, or redeem with the issuer or fund, only if redemption exists. Important distinction: a marketplace is not a buyer, and a buyback is not cash unless owed under contract and properly funded https://joviantan.com/insights/who-controls-your-crypto-multisig-mpc-custody https://joviantan.com/images/insights/who-controls-your-crypto-multisig-mpc-custody.png Who Controls Your Crypto? Illustration titled Who Controls Your Crypto? Multisig, MPC, Custody Risk. A gold Bitcoin coin sits inside an open bank vault, while a gold key on a pedestal is linked by lines of light to three glowing people icons https://joviantan.com/images/insights/custody-control-stack.png The custody control stack: five layers that overlap but are not interchangeable. Diagram headed The digital asset custody control stack, an illustrative framework showing five linked layers. Legal rights: who is entitled? Governance: who changes rules? Signing: who can authorise? Operations: who can pause? Recovery: who restores access? Below: a secure design aligns all five layers, with technical controls, independent governance and tested recovery https://joviantan.com/images/insights/custody-models.png Custody models compared. There is no universal best choice. Comparison headed Custody models: risk moves, it does not disappear. Self-custody: direct key control, owner manages recovery, key-loss and user-error risk. Exchange custody: convenient account access, provider manages keys, counterparty and access risk. Institutional custody: specialist controls may help, contract defines responsibilities, provider and legal risk. Below: ask who controls keys, policy, recovery and legal rights https://joviantan.com/images/insights/custody-multisig-vs-mpc.png Multisig and MPC compared. Conceptual only: implementations vary. Conceptual diagram headed Multisig vs MPC. Multisignature: keys A, B and C feed a 2-of-3 policy requiring two valid signatures under wallet or chain rules. MPC or threshold signing: shares A, B and C feed a threshold protocol in which participants cooperate in a cryptographic signing protocol. Below: neither approach removes governance, implementation or recovery risk https://joviantan.com/images/insights/custody-risk-map.png The custody risk map: evaluate the entire system, not only the wallet technology. Map headed The custody risk map, with digital asset custody (keys, governance, operations, legal rights) at the centre, linked to five risks. Key compromise: stolen key or lost share. Insider abuse: privileged access. Operational failure: outage or failed recovery. Legal or counterparty: insolvency or jurisdiction. Policy concentration: one admin changes rules https://joviantan.com/images/insights/custody-checklist.png The 12-point custody checklist. The questions below explain each point. Checklist headed 12 questions to ask a custody provider, asking for evidence and not just assurances on each: key generation and backups, signing threshold and approvals, separation of duties, withdrawal controls, recovery testing, client asset segregation, lending and reuse permissions, independent security reviews, incident response, subcontractors and jurisdiction, insurance exclusions, exit and portability https://joviantan.com/insights/you-own-the-token-do-you-own-the-asset https://joviantan.com/images/insights/you-own-the-token-do-you-own-the-asset.png You Own the Token, But Do You Own the Asset? Illustration titled You Own the Token. Do You Own the Asset? The Hidden Risks of Real-World Asset Tokenization. A glowing gold coin bearing a building icon sits on a pedestal, linked by lines of light to a modern apartment building and a sealed property deed https://joviantan.com/images/insights/rwa-ownership-chain.png The token, legal documents, issuing structure and underlying asset are connected, but they are not automatically the same thing. Flow diagram headed The legal chain behind an RWA token, with four linked boxes. Blockchain token: records token and transfers. Legal terms and rights: defines what the holder actually owns. Issuer, SPV or trust: owns the asset or owes the contractual duty. Underlying asset: property, bonds, gold, etc. Key question: does a token transfer legally transfer the underlying right? It depends on the structure and governing law https://joviantan.com/images/insights/rwa-insolvency-structures.png Different legal structures can produce very different recovery paths if an issuer fails. Illustrative only. Comparison headed Same asset story, different insolvency outcomes. A, SPV equity: the investor owns company shares; the asset belongs to the SPV and recovery depends on creditors, company law and insolvency rules. B, security or trust: documents create enforceable security or trust rights, which may add protection, subject to valid creation, priority and applicable law. C, issuer promise: the investor holds a contractual claim; without effective security or proprietary rights, the holder may be an unsecured creditor https://joviantan.com/images/insights/rwa-five-layer-test.png The Five-Layer Ownership Test. A framework for due diligence, not a legal opinion or numerical safety score. Framework headed The Five-Layer Ownership Test: an RWA token is only as strong as the chain connecting these layers. 01 Asset existence: does the underlying asset actually exist? 02 Verification and value: can its condition, ownership and value be independently checked? 03 Legal entitlement: what enforceable rights does one token confer? 04 Protection and continuity: what happens if the issuer, custodian or platform fails? 05 Exit and enforcement: can the holder redeem, sell or enforce the claim in practice? https://joviantan.com/images/insights/rwa-liquidity-matrix-v2.png Technical transferability and real-world market liquidity are separate dimensions. Conceptual illustration, not market data. Two-by-two matrix headed Transferability is not the same as liquidity, with technical transferability on the vertical axis and actual market liquidity on the horizontal axis. Transferable, hard to sell: tokens move, but buyers are scarce or discounts are large. Transferable with a market: still assess liquidity under stressed conditions. Restricted and hard to exit: transfer limits combine with few willing buyers. Restricted, but with a market: transfer limits still need separate checking https://joviantan.com/insights/can-crypto-be-frozen https://joviantan.com/images/insights/can-crypto-be-frozen.png Can Crypto Be Frozen? The Hidden Controls Behind Digital Assets Illustration titled Can Crypto Be Frozen? Ethereum, Bitcoin, Tether and USDC coins encased in blocks of ice, with a judge's gavel and a classical courthouse behind them https://joviantan.com/images/insights/can-crypto-be-frozen-freezeability.png Network-level design, token-level controls and custodial restrictions are different things. Diagram headed Crypto Freezeability, comparing four cards. Native BTC: no central issuer with a standard token blacklist; control point, wallet keys and legal process. Native ETH: no central Ethereum account-freeze switch; control point, wallet keys and ecosystem. USDT and USDC: issuer may block supported token addresses; control point, issuer contract controls. Exchange balance: custodian may restrict withdrawals or accounts; control point, platform and legal process. https://joviantan.com/images/insights/can-crypto-be-frozen-enforcement.png Tracing and identifying a person is different from having the power and means to freeze an asset. Six-step process diagram headed From Transaction Trace to Possible Freeze: report or intelligence; trace on-chain transactions; corroborate identity and facts; find an issuer, exchange or custodian; use available legal process; restrict, seize, or keep tracing. A note says a freeze is not automatic. https://joviantan.com/images/insights/can-crypto-be-frozen-fund-flow.png An intermediary or token issuer may be a useful intervention point, but one is not always available. Flow diagram headed Following Funds Across Services and Chains: suspected illicit funds, exchange or off-ramp, self-custody wallet, mixer or privacy tool, cross-chain bridge, another chain or service. Two possible intervention points are marked: exchange account controls at the exchange, and a token issuer blocklist at the wallet. https://joviantan.com/insights/beyond-binance-hack-proof-of-reserves https://joviantan.com/images/insights/beyond-binance-hack-proof-of-reserves.png Beyond the Binance Hack: What Proof of Reserves Can and Cannot Prove Illustration titled Beyond the Binance Hack: What Proof of Reserves Can and Cannot Prove. A magnifying glass shows a reserves panel with a tick and a liabilities panel with question marks, in front of an open vault of coins beneath a Binance sign, with checklists headed Proof of Reserves and Proof of Liabilities https://joviantan.com/insights/binance-security-crisis-customer-trust https://joviantan.com/images/insights/binance-crisis-of-trust.png Beyond the Hack: Can Binance Turn a Security Crisis into a New Standard for Trust in Crypto? Illustration titled The Hack: Crisis of Trust? The Binance logo sits above the headline and the words What It Really Means for Your Assets, and the New Standard for Exchange Trust, beside a cracked padlock bearing the Binance logo, a hooded figure and a candlestick chart https://joviantan.com/gallery https://joviantan.com/images/jovian-tan/2026-10-a12-aelion-twelve-marriott-pj.jpg Presenting at the Aelion Twelve (A12) Asia-Pacific DAO Co-builders Conference. (On stage · October 2026 · Marriott Petaling Jaya) Dr. Jovian Tan speaking at a lectern in front of a presentation slide about GPU clusters https://joviantan.com/images/jovian-tan/keynote-cloud-slide.jpg Delivering a keynote on cloud technology in Kuala Lumpur. (On stage) Dr. Jovian Tan on stage with a microphone, gesturing towards a slide showing a network diagram https://joviantan.com/images/jovian-tan/speaking-conference-audience.png Speaking from the floor during a conference session in China. (On stage) Dr. Jovian Tan in a dark blazer and scarf speaking into a microphone, standing among a seated conference audience https://joviantan.com/images/jovian-tan/2017-08-speaking-microphone.jpg Addressing an audience in Beijing. (On stage · August 2017) Dr. Jovian Tan in a grey suit speaking into a microphone beside a lectern https://joviantan.com/images/jovian-tan/2018-thailand-high-frequency-trading-platform.jpg Presenting a next-generation high-frequency trading platform in Thailand. (On stage · 2018 · Thailand) Dr. Jovian Tan in a navy three-piece suit with a rose boutonnière, presenting on stage in front of a screen listing trading bots and asset investment https://joviantan.com/images/jovian-tan/2019-blockchain-press-conference.jpg At a press conference on future applications of blockchain. (On stage · 2019) Dr. Jovian Tan answering questions at a press conference, facing microphones from several Chinese media outlets https://joviantan.com/images/jovian-tan/cryptocurrency-expo-panel.jpg Speaking on a panel at a cryptocurrency expo. (Industry & events) Dr. Jovian Tan speaking on a panel, seated in an armchair with a microphone https://joviantan.com/images/jovian-tan/brandlaureate-smes-bestbrands.jpg At The BrandLaureate SMEs BestBrands event. (Industry & events) Dr. Jovian Tan in a three-piece suit in front of a BrandLaureate SMEs BestBrands backdrop https://joviantan.com/images/jovian-tan/2026-10-a12-aelion-twelve-front-row.png Seated in the front row at the Aelion Twelve (A12) conference. (Industry & events · October 2026 · Marriott Petaling Jaya) Dr. Jovian Tan seated in the front row of a conference audience https://joviantan.com/images/jovian-tan/forbes-office.jpg At the Forbes office in New York. (Industry & events) Dr. Jovian Tan standing beside the Forbes logo on an office wall https://joviantan.com/images/jovian-tan/2018-israel-china-ai-blockchain-partnership.jpg Special guest at an Israel–China innovation partnership event on next-generation AI and blockchain development.Photograph: ChainHoo Finance (链虎财经) (Industry & events · 2018) Dr. Jovian Tan in a navy three-piece suit wearing a guest badge, seated and listening in a conference audience https://joviantan.com/images/jovian-tan/2017-brandlaureate-bestbrands-awards.jpg Attending The BrandLaureate BestBrands Awards. (Industry & events · 2017) Dr. Jovian Tan in a black tuxedo and bow tie in front of The BrandLaureate BestBrands Awards 2016–2017 backdrop https://joviantan.com/images/jovian-tan/2014-microsoft-singapore.png At Microsoft’s Singapore office, during his years with the company. (Industry & events · 2014 · Singapore) Dr. Jovian Tan smiling in a red sweater with a staff lanyard and notebook, standing in a corridor at a Microsoft office https://joviantan.com/images/jovian-tan/microsoft-headquarters-seattle.jpg At Microsoft’s headquarters near Seattle. (Industry & events · Seattle area, USA) Dr. Jovian Tan in a black puffer jacket, smiling beside a large Microsoft logo sign on a tree-lined campus https://joviantan.com/images/jovian-tan/2018-blockchain-centre-shanghai.jpg Signing the guest wall at Blockchain Centre Shanghai, as an invited guest at a global blockchain event. (Industry & events · 2018 · Shanghai, China) Dr. Jovian Tan smiling as he signs a guest wall branded Blockchain Centre Shanghai https://joviantan.com/images/jovian-tan/great-wall-of-china.jpg Visiting the Great Wall of China. (Beyond work · Beijing, China) Dr. Jovian Tan smiling in front of the Great Wall of China winding across green hills https://joviantan.com/images/jovian-tan/2017-08-beijing-tiananmen.jpg In Tiananmen Square, Beijing. (Beyond work · August 2017 · Tiananmen Square, Beijing, China) Dr. Jovian Tan smiling in Tiananmen Square, Beijing, with a traditional gate building in the background https://joviantan.com/images/jovian-tan/2013-playing-guitar.jpg Playing guitar during his student years at Universiti Teknologi Malaysia (UTM). (Beyond work · 2013) Dr. Jovian Tan smiling as he plays an acoustic guitar, seated on garden steps among tropical plants and red ginger flowers https://joviantan.com/images/jovian-tan/2013-rafflesia.jpg Beside a Rafflesia in bloom at Gunung Gading National Park, Sarawak. (Beyond work · 2013) Dr. Jovian Tan in a cap and white T-shirt, crouching beside a large red Rafflesia flower in bloom on a rainforest floor https://joviantan.com/images/jovian-tan/2014-new-york-wall-street.png Visiting Wall Street in New York. (Beyond work · 2014 · New York, USA) Dr. Jovian Tan in sunglasses and a striped T-shirt, carrying a shoulder bag and looking up beneath a tree among city buildings https://joviantan.com/images/jovian-tan/2016-tandem-skydive.jpg Tandem skydiving from 14,000 feet in Australia. (Beyond work · 2016) Dr. Jovian Tan in freefall during a tandem skydive, grinning with both thumbs up, farmland and a river far below https://joviantan.com/images/jovian-tan/christmas-carols-piano-kuala-lumpur2.png Playing piano in a Christmas carol performance in Kuala Lumpur. (Beyond work) Dr. Jovian Tan playing a grand piano beside a choir dressed in black, in an atrium hung with golden Christmas lights https://joviantan.com/contact https://joviantan.com/images/jovian-tan/speaking-warm-conference-event.png Dr. Jovian Tan in a dark blazer and black scarf speaking into a microphone at a conference, with a seated audience and warm golden curtains behind him https://joviantan.com/zh/insights/who-controls-the-exit-crypto-exchanges-gatekeepers https://joviantan.com/images/insights/who-controls-the-exit-crypto-exchanges-gatekeepers-zh.png 谁掌控资金的出入口？当加密资产交易所成为资金守门人 插图，标题为加密交易所正在成为你资金的守门人？上方写有“币安 限制 MEXC”，下方写有“币安–MEXC 交易限制、160亿美元监管调查，对加密用户意味着什么？”一串金色代币沿桥梁从币安关口驶向 MEXC 关口，被标有“转账受限 交易被阻止”的红色发光屏障拦下 https://joviantan.com/images/insights/exchange-exit-key-dates-zh.png 一项政府命令、一份已刊登的规则，以及一项独立的平台决定。 六张卡片，标题为“币安、MEXC 与 FinCEN：关键日期”。2026年11月6日：FinCEN 根据第9714(a)条宣布命令，涉及俄罗斯非法金融。2026年11月7日：MEXC 回应称正审阅命令，平台运作正常。2026年11月10日：命令在《联邦公报》正式刊登，91 FR 72420。2026年11月10日：币安实施限制，拒绝转出并扣留入金审查。2026年12月10日（突出显示）：命令对受涵盖的美国金融机构生效。属指控，并非裁决：至少160亿美元，为美国财政部对2021年5月至2026年5月的估算 https://joviantan.com/images/insights/exchange-exit-access-layers-zh.png 概念示意，并非特定交易路径。 示意图，标题为“准入在哪里受到控制”，为加密资产到法定货币流程的简化示意。一排四道关卡，并注明谁可以拒绝。公共区块链，负责转移并结算代币：协议规则。交易所或托管机构（突出显示），负责入账、扣留或拒绝：交易所合规。稳定币发行商，负责铸造与赎回代币：发行商控制。银行或法定货币通道，负责兑换为银行货币：银行及监管机构。下方：协议开放，并不代表任何中介机构都必须接受转账 https://joviantan.com/zh/insights/the-new-financial-order https://joviantan.com/images/insights/the-new-financial-order-zh.png 全球金融新秩序：区块链是否正在悄然重塑全球银行体系？ 插图，标题为全球金融新秩序：区块链是否正在悄然重塑全球银行体系？并写有“从代币化存款到稳定币，银行正加速走向区块链。”一座标有“银行”的古典建筑矗立在发光的地球与城市高楼旁，一串发光的区块横跨世界地图 https://joviantan.com/images/insights/agora-real-value-test-zh.png 公布的测试数据，并非全市场采用率。 六张卡片，标题为“Agorá 项目：2026年7月真实价值测试”，为 BIS 公布的数据，属于受控测试，并非商业化基准。参与机构：28家金融机构及中央银行。交易场景：17种场景，以指定货币完成30笔交易。交易总额：约80万瑞士法郎。金额范围：单笔9,000至125,000瑞士法郎或等值当地货币。速度：从支付发起到结算平均约80秒。范围：受控测试，原型尚未与实时全额结算系统整合。来源：BIS 创新中心，Agorá 项目页面，2026年11月9日核对 https://joviantan.com/images/insights/digital-finance-architecture-zh.png 概念示意。不同平台的设计、治理模式和结算资产各不相同。 示意图，标题为“数字金融的新兴架构”，为概念模型，共五层。用户：个人、企业及金融机构。支付与结算：国内、跨境及批发。代币化货币（突出显示）：代币化存款、受监管稳定币、批发型央行数字货币。共享账本基础设施：共享状态、可编程性、互操作性。监管与信任：中央银行、监管机构、法律与合规。附注：无论记录在哪个账本上，每种货币形式所代表的法律债权都不相同 https://joviantan.com/images/insights/three-forms-of-digital-money-zh.png 账本相似，法律债权不同。 三栏图，标题为“三种数字货币”：相似的数字界面，可能代表截然不同的法律债权。代币化存款：由商业银行发行，是对该银行的债权；主要风险为银行信用、流动性和运营；关键问题是银行是否稳健。受监管稳定币：由私人机构发行，权利由法律和条款界定；主要风险为储备、赎回和托管；关键问题是持有人能否按面值赎回。批发型央行数字货币：由中央银行发行，是对中央银行的债权；主要风险为设计、准入和政策；关键问题是谁可以使用。下方：法律债权比它所记录的账本更重要 https://joviantan.com/images/insights/tokenised-cross-border-settlement-zh.png 哪些改变了，哪些没有改变。 流程图，标题为“跨境支付：代币化改变了什么”，为简化示意，共三行。代理行模式：汇款银行发送支付报文，代理行各自记账并受截止时间限制，收款银行事后对账。共享账本模式：代币化存款与储备在可编程平台上移动，条件写入智能合约，原子结算使所有环节全部完成或全不完成。两种模式仍然需要：KYC、制裁筛查、反洗钱等合规检查；流动性与外汇资金；以及由法律而非代码决定的法律终局性。下方：代币化改变的是支付的协调方式，而不是背后的金融义务 https://joviantan.com/zh/insights/ueno-bank-collapse-crypto-infrastructure https://joviantan.com/images/insights/ueno-bank-collapse-crypto-infrastructure-zh.png 乌埃诺银行倒闭：这对加密基础设施究竟意味着什么？ 插图，标题为乌埃诺银行倒闭：这对加密基础设施究竟意味着什么？一座标有 Ueno Bank 的古典银行建筑出现裂缝，隔着一座桥与装有区块链方块的发光玻璃金库相对，背景是城市天际线与上升的K线图 https://joviantan.com/images/insights/ueno-deposit-guarantee-zh.png 央行公告时的法定保障上限。 示意图，标题为“巴拉圭存款保障：上限意味着什么”，未按比例绘制。一条横条在上限处分开：受保障的合资格存款，通过存款保障机制支付；超过上限的余额，取决于银行处置程序。上限为2.283亿瓜拉尼，即75个月最低工资，按每名存款人、每家机构计算，而不是按账户计算；这并不代表可以立即提取，也不保证超过上限的部分能够全数追回。来源：巴拉圭中央银行，2026年11月3日 https://joviantan.com/images/insights/ueno-key-facts-zh.png 监管机构与评级机构已公布的内容。 六张卡片，标题为“乌埃诺银行：关键事实”。决定：2026年11月2日巴拉圭央行董事会决定启动处置。触发条件：财务报表调整后偿付能力指标低于法定最低要求的50%。评级：FIX SCR 于2026年11月3日由 AA-py 降至 Dpy。关注问题：关联方交易，以及资产质量与文件记录。保障上限：每名存款人、每家机构2.283亿瓜拉尼。现状：处置程序进行中，超过上限部分的追回情况未定。来源：巴拉圭中央银行，2026年11月3日及5日；FIX SCR，2026年11月3日 https://joviantan.com/images/insights/ueno-risk-comparison-zh.png 定性分析框架，并非排名。 三栏图，标题为“三种模式，三种风险结构”。吸收存款的银行：银行对存款负有债务，资金在资产负债表上运用；资本、流动性与监督等审慎监管；合资格存款设有上限的存款保障。中心化加密平台：平台控制私钥，可能对客户负有资产债务；放贷、质押、再抵押等资产再利用风险；保障因地而异，通常没有存款保险。非托管模式：用户控制私钥，没有中介信用风险；交易可在链上独立核验；用户承担遗失、钓鱼、合约漏洞等密钥风险 https://joviantan.com/images/insights/ueno-customer-asset-flow-zh.png 各模式中，控制权与风险敞口位于何处。 流程图，标题为“客户资产流向：银行与加密模式”，共三行。吸收存款的银行：客户存款，银行持有并放贷，贷款与收益带来信用与流动性风险。中心化加密平台：客户转入资产，平台托管并持有私钥，资产可能用于交易或借贷。非托管模式：用户持有私钥，由智能合约执行，链上结算公开且可核验。下方：谁控制资产？谁对客户负有债务？ https://joviantan.com/zh/insights/ai-learning-how-to-hack-cybersecurity-defence https://joviantan.com/images/insights/ai-learning-how-to-hack-cybersecurity-defence-zh.png AI 正在学会如何发动网络攻击：我们准备好了吗？ 插图，标题为“AI 正在学会如何发动网络攻击？我们准备好了吗？”，副标题为“人工智能正在重塑网络犯罪以及网络安全的未来”。一个带有挂锁芯片的发光数字大脑、地球和一排安全盾牌，下方是一名戴兜帽的人面对显示威胁图标的屏幕 https://joviantan.com/images/insights/ai-hack-threat-landscape-zh.png 人工智能既是攻击者的工具，也是新的攻击目标。 六张卡片，标题为“人工智能网络风险的两个方向”：利用 AI 发动攻击（钓鱼、身份冒充、合成媒体及更快速的侦察）；直接攻击 AI 系统（提示词注入、资料泄露、检索投毒及工具滥用）；利用人类信任（紧迫感、权威感、熟悉感与绕过流程）；基础设施暴露（凭证、API、云服务、终端设备与数据存储）；利用 AI 防御（警报筛选、异常检测、调查与响应）；核心原则：核实身份、限制权限，并监控每项关键操作 https://joviantan.com/images/insights/ai-hack-attack-surface-zh.png 模型只是攻击面的一部分。 六张编号卡片，标题为“人工智能应用的潜在攻击面”：输入与提示词（用户提示词、网页、电邮及上传文件）；数据与检索（知识库、向量数据库、权限与来源）；模型与上下文（系统指令、上下文窗口及模型供应链）；工具与 API（电邮、数据库、代码执行及支付集成）；输出与操作（验证、策略检查、审批及审计记录）；运营管理（日志、监控、事件响应与回滚）。下方：模型只是其中一个组件，整个工作流程都必须具备安全控制 https://joviantan.com/images/insights/ai-hack-dbir-initial-access-zh.png 一般数据泄露统计，并非人工智能归因数据。 条形图，标题为“2025年 DBIR 已知初始入侵途径”，采用同一尺度：凭证滥用22%、漏洞利用20%、网络钓鱼16%。说明：并非 AI 专属，属一般数据泄露的入侵途径，并非由 AI 导致的攻击；勒索软件出现在44%的数据泄露中，上一年为32%。来源：Verizon《2025年数据泄露调查报告》 https://joviantan.com/images/insights/ai-hack-eight-layers-zh.png 安全来自一套控制体系，而不是单一产品。 八张编号卡片，标题为“人工智能网络安全八层防御框架”：身份验证（抗钓鱼认证与独立核实渠道）；最小权限（狭窄权限、短期凭证与即时撤销）；审批关卡（付款及高影响操作实施双重控制）；数据保护（分类、加密及权限感知检索）；安全 AI 设计（不可信输入处理与输出验证）；持续监控（日志、异常检测与响应责任）；安全测试（红队演练、提示词注入与权限测试）；恢复能力（备份、遏制及代理停用程序） https://joviantan.com/images/insights/ai-hack-incident-response-zh.png 在事件发生前，先确定由谁、以何种方式采取行动。 流程图，标题为“实用的网络安全事件响应流程”，共六步：发现异常访问、工具调用、资料导出或付款请求；遏制，暂停受影响的代理、账户、集成服务或终端；撤销凭证、令牌、会话及 API 密钥；调查，保存日志、确认影响范围、追踪操作与资料；恢复可信系统并验证完整性；改进，检讨根本原因、更新控制并再次演练。下方：在真正需要之前，先演练整个流程 https://joviantan.com/zh/insights/ai-deepfakes-financial-trust-infrastructure https://joviantan.com/images/insights/ai-deepfakes-financial-trust-infrastructure-zh.png 你还能相信自己的眼睛吗？ 插图，标题为“你还可以相信自己的眼睛吗？”，副标题为“AI 深度伪造如何冲击全球信任基础设施”。一名男子的脸从中间分开，一半是真实皮肤，另一半是发光的数字线框，旁边是显示银行图标的笔记本电脑和成堆硬币 https://joviantan.com/images/insights/deepfake-attack-chain-zh.png 合成媒体如何变成资金损失。 图表标题为“深度伪造攻击链”，分为五步：合成声音、面孔或身份资料；冒充可信人物或机构以建立信任；以权威、保密与时间压力制造紧迫感；诱导受害者绕过正常控制；最终造成转账、账户接管或数据泄露等资金损失。下方说明：外表可信，并不代表身份真实或授权有效 https://joviantan.com/images/insights/deepfake-fbi-ai-complaints-zh.png 四项独立指标，并非同一尺度。仅为美国投诉数据。 四张独立数字卡片，标题为“2025年向 FBI 报告的 AI 相关网络犯罪”：涉及人工智能相关信息的投诉22,364宗；上述投诉的调整后损失8.93亿美元；报告存在 AI 关联的投资诈骗损失超过6.32亿美元；涉及 AI 的就业诈骗损失约1,300万美元。说明：单位与类别各不相同，数字可能重叠，不应相加。来源：FBI 互联网犯罪投诉中心《2025年互联网犯罪报告》 https://joviantan.com/images/insights/deepfake-payment-workflow-zh.png 每一步核验的内容各不相同。 图表标题为“更安全的大额付款流程”，分为五步：接收指令，不要仅凭视频或电邮；使用预先登记的号码独立回拨；核对收款方，确认收款账户及变更记录；由两名不同人员双人审批；执行与留档，设定限额并保留审计记录。下方说明：任何核验未通过，应暂停付款、升级处理并保留证据 https://joviantan.com/images/insights/deepfake-trust-layers-zh.png 五层相互独立的金融信任。 图表标题为“五层金融信任架构”，每层列出核心问题与控制示例：身份（对方是谁？身份核验及凭证检查）、渠道（这是否为获批准的沟通渠道？预先登记联系方式及安全门户）、权限（此人能否批准该操作？基于角色的权限及双重控制）、交易（具体批准的内容是什么？签署交易细节及收款方核验）及证据（能否还原整个决策过程？防篡改日志及事件记录） https://joviantan.com/zh/insights/financial-system-24-7-bank-run-liquidity-risk https://joviantan.com/images/insights/financial-system-24-7-bank-run-liquidity-risk-zh.png 金融系统正在迈向全天候时代，但它准备好应对挤兑了吗？ 插图，标题为“金融系统永不休市，但它准备好应对挤兑了吗？”，副标题为“全天候金融下，流动性、信任与监管能否跟上？”。地球被时钟般的箭头环绕，旁边是一座立柱式银行建筑，下方的玻璃面板展示硬币、房屋、金条与证书 https://joviantan.com/images/insights/bank-run-transaction-chain-zh.png 全天候市场是一条依赖链。 图表标题为“全天候金融交易链”：五个相连环节，分别为交易（订单与定价）、清算（核对双方义务）、结算（资金与资产完成转移）、托管（记录资产所有权）及赎回（转换为可用资金）。下方说明持续开放不等于持续承载能力 https://joviantan.com/images/insights/bank-run-svb-withdrawals-zh.png 约八小时内，接近四分之一的存款被要求提取。 柱状图标题为“硅谷银行：数字化挤兑的速度”。存款总额约1,660亿美元，2023年3月9日约八小时内的提款请求约420亿美元。说明：接近四分之一的存款，约25%。来源：加州金融保护与创新部，2023年5月8日 https://joviantan.com/images/insights/bank-run-liquidity-mismatch-zh.png 代币、资产与现金的运行速度并不相同。 图表标题为“代币化资产的流动性错配”，分为三层。代币层：全天候流转，可能支持持续转让与交易。资产层：变现较慢，估值较慢、买家有限，变现周期较长。现金层：受银行时间约束，包括银行营业时间、支付截止时间、流动性限制与赎回条件。下方说明可转让性并不保证即时流动性 https://joviantan.com/images/insights/bank-run-liquidation-spiral-zh.png 自动清算如何自我强化。 图表标题为“清算连锁反应如何形成”，展示杠杆化 DeFi 市场中可能出现的反馈循环：价格下跌、抵押品走弱、智能合约触发清算、抵押品被卖入下跌或深度不足的市场、价格再跌，并可能触发下一轮清算。下方说明严重程度取决于抵押品、预言机、市场深度与协议设计 https://joviantan.com/images/insights/bank-run-resilience-layers-zh.png 全天候金融体系必须守住的七个层面。 七张卡片，标题为“全天候金融韧性的七个层面”：资产质量（透明、高质量资产与合理估值）、流动性（经压力测试的缓冲与可信赎回）、托管（资产隔离、控制、恢复与独立监督）、结算（法律最终性与安全结算资产）、技术（网络安全、冗余、监控与事件响应）、治理（明确问责与经过测试的应急权力）及监管（跨境合作与数据共享） https://joviantan.com/zh/insights/when-ai-starts-moving-money-who-is-accountable https://joviantan.com/images/insights/when-ai-starts-moving-money-who-is-accountable-zh.png 当人工智能开始掌控资金：谁来为它的金融决策负责？ 插图，标题为“当人工智能开始掌控资金：谁来为它负责？智能金融、自动化支付，以及人工智能驱动资金流动的新风险”。一个人形 AI 手托发光的数字钱包，旁边是地球、比特币、泰达币与美元硬币、显示付款执行中的界面、安全盾牌及银行图标 https://joviantan.com/images/insights/ai-payments-boe-autonomy-zh.png 部署已很普遍，广泛自主仍属少数。仅限调查受访者。 柱状图标题为“英国金融服务业的 AI 智能体自主程度”：在已部署 AI 智能体的受访机构中，80%用于决策支持，46%为受限执行，22%为条件自主，2%为广泛自主。另有说明：即使计入未来12个月的计划部署，预计把智能体用于付款等指定高影响职能的受访机构也不超过9%。来源：英格兰银行与英国金融行为监管局，2026年10月 https://joviantan.com/images/insights/ai-payments-three-layers-zh.png 智能体支付的三个层次。 图表标题为“智能体支付的三个层次”。意图与规划：理解用户目标，生成结构化付款请求。授权与控制：独立核验身份、范围、收款方、金额及审批。结算与记录：由银行、卡组织、支付服务商或区块链执行并记录。下方说明模型提出建议，独立控制授权，基础设施负责结算 https://joviantan.com/images/insights/ai-payments-risk-map-zh.png 智能体支付风险集中在哪些环节。 六张卡片，标题为“AI 驱动资金流动的六类风险”：操作风险（错误决策、错误收款方或意外交易）、安全风险（提示词注入、账户接管、密钥或数据泄露）、金融风险（未经授权的付款、损失或流动性冲击）、法律与监管风险（责任归属、消费者保护及合规义务不清）、治理风险（权限过大、监督不足或审计链断裂）及声誉风险（客户信任受损、争议与品牌影响）。下方说明控制重点为预防、限制、监测、暂停及追责 https://joviantan.com/images/insights/ai-payments-secure-architecture-zh.png 更安全的自主支付架构。 图表标题为“更安全的 AI 资金流动架构”，分为三层。应用层：AI 智能体理解意图、分析选项并生成付款指令。控制层：身份与政策核验主体、权限、限额、收款方、风险与审批。执行层：通过银行、卡组织、支付服务商或链上系统执行并留痕。必须具备：明确的授权主体、交易限额与范围、实时监控与可撤销，以及审计记录与责任归属 https://joviantan.com/zh/insights/stablecoins-new-financial-infrastructure https://joviantan.com/images/insights/stablecoins-new-financial-infrastructure-v2-zh.png 稳定币：新一代金融基础设施？ 插图，标题为“稳定币：下一代的全球金融基础设施？USDT、USDC 与代币化货币如何重塑全球支付体系”。USDT 与 USDC 硬币立于发光的地球与银行建筑前，旁边的面板分别标有全球支付、跨境结算、RWA 资产代币化及可编程货币 https://joviantan.com/images/insights/stablecoin-trust-stack-zh.png 稳定币信任架构：每一层都必须稳固。 图表标题为“稳定币信任架构”，显示支付承诺所依赖的六个层面：发行人治理、储备质量与托管、法律权利与赎回、银行渠道、区块链与智能合约，以及钱包及中介控制。下方说明任何一层出现弱点，都可能削弱整体支付承诺 https://joviantan.com/images/insights/stablecoin-market-share-zh.png 估算市场份额，已四舍五入。不同来源和日期的估算会有差异。 柱状图标题为“两大发行人占据主要市场份额”：USDT 与 USDC 合计约占稳定币市场的90%，其他发行人约占10%。来源：IMF 部门文件第2026/021号，2026年10月，已四舍五入 https://joviantan.com/images/insights/stablecoin-lifecycle-zh.png 简化运作流程。实际发行与赎回取决于发行人的条款。 图表标题为“储备支持型稳定币的运作流程”，包含五个相连步骤：法币入金，客户向发行人转入资金；铸造代币，发行人创建代币；链上转账，在钱包之间转移；申请赎回，合资格持有人提交代币；法币付款，由发行人或合作方付款。铸造及赎回受发行条款、资格审核与银行渠道影响 https://joviantan.com/images/insights/stablecoin-reserve-model-zh.png 储备模式简化示意。储备构成及法律权利因发行人而异。 图表标题为“储备模式：什么在支持代币价值？”。发行人创建及赎回代币；储备资产池持有现金、存款及短期国债；托管机构与银行负责资产保管及资金渠道；持有人享有合同约定权利。信任需要储备质量、法律权利及可信赎回机制共同支撑。实际储备构成及法律权利因发行人而异 https://joviantan.com/images/insights/stablecoin-payment-rails-zh.png 稳定币轨道与传统代理银行体系比较。 比较图标题为“两种支付轨道，各有不同取舍”。稳定币轨道：通常全天候、链上确认、智能合约逻辑、出金或商户接受。传统代理银行体系：受银行营业及截止时间限制、多家银行参与处理、依赖传统系统整合、本地银行渠道。端到端成本、合规及兑换体验，比单纯速度更重要 https://joviantan.com/images/insights/stablecoin-risk-map-zh.png 稳定币风险集中在哪些环节。 风险图标题为“稳定币风险集中在哪些环节？”。发行人与资产层：储备与流动性、赎回准入、银行或托管机构。技术与市场层：智能合约或区块链、冻结或治理、脱锚或市场。风险控制应覆盖发行人、中介机构、钱包及区块链网络 https://joviantan.com/images/insights/stablecoin-future-money-zh.png 三种可能并存的数字货币形式。 图表标题为“未来金融体系很可能是混合式架构”。稳定币（私人机构发行的债权）、代币化存款（商业银行负债）及央行货币（中央银行负债）都连接到互联互通的金融轨道，用于支付、结算及资金管理 https://joviantan.com/zh/insights/tokenisation-liquidity-illusion-rwa-secondary-markets https://joviantan.com/images/insights/tokenisation-liquidity-illusion-rwa-secondary-markets-zh.png 流动性幻觉 插图，标题为“流动性幻觉：资产上链就能轻松变现？揭开 RWA 代币化背后的流动性真相”。一块标有“RWA 代币化”的玻璃方块内是一栋现代公寓楼，下方两座悬崖分别标有“资产代币化”和“真实流动性”，中间是断开的桥和问号 https://joviantan.com/images/insights/liquidity-test-zh.png RWA 流动性测试。这是示意框架，不是评分模型。 图表标题为“RWA 流动性测试”，这是示意框架，列出投资者完成退出必须通过的五个环节。法律权益：代币能否依法转让？买家准入：谁有资格购买？市场深度：是否存在真实买盘？成交质量：以什么价格成交多少？结算确定性：退出能否真正完成？下方说明可转让性不等于流动性，任何一个环节存在弱点，都可能削弱退出能力 https://joviantan.com/images/insights/liquidity-study-sample-zh.png Swinkels（2023）的样本：属于历史数据，规模有限，并集中于一座城市。 柱状图标题为“研究中 58 处物业的地理分布”：52 处位于底特律，6 处位于美国其他地区。来源：Swinkels（2023），2019年10月至2021年2月期间代币化的 58 处美国住宅出租物业 https://joviantan.com/images/insights/liquidity-research-evidence-zh.png 两项不同研究的部分发现。这些数据不能直接互相比较。 三组数字，标题为“两项不同研究的部分发现”。Swinkels（2023）：每处物业的持有人中位数为 254，所有权变更的平均间隔约为一年。BIS 工作论文 1311：自然灾害后两天内交易量累计上升 35%，仅适用于所研究的美国平台数据及回购机制。这些数据不能直接互相比较 https://joviantan.com/images/insights/liquidity-exit-routes-zh.png 三条路径，交易对手、条件和风险各不相同。 图表标题为“投资者实际如何退出代币化资产”。发行方创建代币，投资者 A 通过一级认购买入。投资者 A 可以出售给合资格的二级市场买家投资者 B，也可以在设有赎回安排时向发行方或基金赎回。重要区别：交易平台不等于买家；回购也不等于现金，除非合同规定应当支付，并且资金到位 https://joviantan.com/zh/insights/who-controls-your-crypto-multisig-mpc-custody https://joviantan.com/images/insights/who-controls-your-crypto-multisig-mpc-custody-zh.png 谁真正掌控你的加密资产？ 插图，标题为“谁掌控你的加密资产？多重签名、MPC、托管风险”。一枚金色比特币位于敞开的金库内，一把发光的钥匙以光线连接三个带勾选标记的人物图标 https://joviantan.com/images/insights/custody-control-stack-zh.png 数字资产托管控制链：五个层面彼此有关，却不能混为一谈。 示意框架图，标题为“数字资产托管控制链”，展示五个相连的层面。法律权利：谁依法享有权利？治理权限：谁能修改规则？签名权限：谁能授权交易？日常运营：谁能暂停转账？恢复机制：谁能恢复访问？下方写明：安全的设计要让五个层面协调一致，包括技术控制、独立治理和经过测试的恢复机制 https://joviantan.com/images/insights/custody-models-zh.png 不同托管模式的比较。没有放之四海而皆准的最佳选择。 比较图，标题为“不同托管模式：风险只是转移，并未消失”。自托管：直接控制密钥，资产所有人负责恢复，存在密钥遗失与操作失误风险。交易所托管：账户操作较方便，服务商管理密钥，存在交易对手与访问风险。机构托管：专业控制措施可能有帮助，合同界定责任，存在服务商与法律风险。下方写明：应询问谁控制密钥、政策、恢复机制和法律权利 https://joviantan.com/images/insights/custody-multisig-vs-mpc-zh.png 多重签名与 MPC 的比较。仅为概念示意，具体实现各不相同。 概念图，标题为“多重签名与 MPC”。多重签名：密钥 A、B、C 连接到 2-of-3 规则，需要两个有效签名。MPC／门限签名：份额 A、B、C 连接到门限协议，多个参与方通过密码学签名协议协作。下方写明：两种方案都不能消除治理、实现与恢复风险 https://joviantan.com/images/insights/custody-risk-map-zh.png 数字资产托管风险地图：应评估整个系统，而不只是钱包技术。 风险地图，标题为“数字资产托管风险地图”，中央是数字资产托管（密钥、治理、运营、法律权利），连接五类风险。密钥泄露：密钥被盗或份额遗失。内部滥用：特权访问。运营故障：服务中断或恢复失败。法律与交易对手风险：破产与司法管辖区。权力集中：单一管理员修改规则 https://joviantan.com/images/insights/custody-checklist-zh.png 12 项托管审查清单。下文逐项说明。 清单图，标题为“向托管服务商提出的 12 个问题”，每项都要看证据，不只听保证：密钥生成与备份、签名门槛与审批、职责分离、提款控制、恢复测试、客户资产隔离、借出与再利用权限、独立安全审查、事件响应、分包商与司法管辖区、保险除外责任、退出与迁移 https://joviantan.com/zh/insights/you-own-the-token-do-you-own-the-asset https://joviantan.com/images/insights/you-own-the-token-do-you-own-the-asset-zh.png 你持有代币，就真的拥有底层资产吗？ 插图，标题为“你持有代币 就拥有资产？RWA 代币化的隐性风险”。一枚刻有立方体图案的发光金币立于底座上，以光线连接一栋现代住宅楼和一份盖有火漆印章的产权文件，背景是吉隆坡双子塔的城市天际线 https://joviantan.com/images/insights/rwa-ownership-chain-zh.png 代币、法律文件、发行结构与底层资产彼此相关，但并不自动等同。 流程图，标题为“代币与底层资产之间的权利链”，由四个相连的方框组成。区块链代币：记录持有人与转移。法律文件与权利：定义代币持有人实际拥有什么。发行方、SPV 或信托：持有资产或承担合同义务。底层现实资产：房地产、债券、黄金等。关键问题：代币转移是否依法转移了底层权益？答案取决于具体结构与适用法律 https://joviantan.com/images/insights/rwa-insolvency-structures-zh.png 不同法律结构在发行方倒闭时，可能产生截然不同的追偿路径。仅作示意。 比较图，标题为“同样的资产宣传，不同的破产结果”。A，SPV 股权：投资者持有公司股份；资产属于 SPV，回收受债权人顺位及公司法影响。B，担保或信托安排：文件建立可执行的担保或信托权益，可能提供额外保护，取决于有效设立、优先顺位与适用法律。C，发行方承诺：投资者只有合同请求权；若无有效担保或资产权益，可能成为无担保债权人 https://joviantan.com/images/insights/rwa-five-layer-test-zh.png 五层所有权测试。尽职调查框架，不构成法律意见，也不是数字化安全评分。 框架图，标题为“五层所有权测试（Five-Layer Ownership Test）”：RWA代币是否可靠，取决于连接这五个层面的链条。01 资产是否真实存在：底层资产是否真实存在？02 核验与估值：资产的状况、所有权和价值能否独立核验？03 法律权益：一枚代币赋予什么可依法执行的权利？04 资产保护与业务连续性：发行方、托管人或平台倒闭时会发生什么？05 退出与执行权利：持有人实际上能否赎回、出售或执行权利？ https://joviantan.com/images/insights/rwa-liquidity-matrix-v2-zh.png 技术上的可转移性与现实市场流动性是两个不同维度。概念示意，不代表市场数据。 二乘二矩阵图，标题为“可转移，不代表有流动性”，纵轴为技术上可转移，横轴为实际市场流动性。可转移但难卖：代币可发送，但买家少或折价大。较易转移且有市场：仍需关注压力时期的流动性。转移受限且难退出：转让限制与买家不足叠加。转移受限但有市场：转让限制仍需独立核查 https://joviantan.com/zh/insights/can-crypto-be-frozen https://joviantan.com/images/insights/can-crypto-be-frozen-zh.png 加密货币可以被冻结吗？数字资产背后的隐形控制机制 插图，标题为“加密货币可以被冻结吗？”。以太坊、比特币、泰达币和 USDC 硬币被封在冰块中并以铁链和挂锁锁住，背后是法槌和一座古典风格的法院建筑 https://joviantan.com/images/insights/can-crypto-be-frozen-freezeability-zh.png 原生网络、代币合约与托管账户的控制机制并不相同。 图表，标题为“加密资产的可冻结性是一条光谱”，比较四类资产。原生 BTC：没有中央发行方提供标准化地址黑名单，主要控制点为私钥持有人和法律程序。原生 ETH：以太坊基础网络没有中央账户冻结开关，主要控制点为私钥持有人和生态服务。USDT 和 USDC：发行方可能封锁特定地址的代币转账，主要控制点为发行方合约权限。交易所余额：平台可能限制账户或暂停提现，主要控制点为托管平台和法律程序。 https://joviantan.com/images/insights/can-crypto-be-frozen-enforcement-zh.png 链上追踪、身份识别、法律权限与实际执行是不同环节。 六步流程图，标题为“链上调查如何走向可能的资产冻结”：收到线索或报案；分析链上交易路径；核实身份与相关证据；找到发行方、交易所或托管方；依法提出请求或申请命令；限制、扣押或继续追查。图中注明冻结并非自动发生。 https://joviantan.com/images/insights/can-crypto-be-frozen-fund-flow-zh.png 交易所和代币发行方可能提供介入点，但并非每条资金路径都有这些节点。 资金流向图，标题为“追踪资金：中途可能出现哪些介入点？”：可疑资金、交易所或法币出入口、自托管钱包、混币器或隐私工具、跨链桥、其他链或服务。图中标出两个可能的介入点：交易所账户限制，以及代币发行方封锁地址。 https://joviantan.com/zh/insights/beyond-binance-hack-proof-of-reserves https://joviantan.com/images/insights/beyond-binance-hack-proof-of-reserves-zh.png 币安安全事件之后：储备证明能证明什么，不能证明什么 插图，标题为“不只是看钱包余额！交易所的储备证明到底能证明什么？”。放大镜下是打勾的储备资产面板和标有问号的用户负债面板，背后是币安标志下一座敞开、堆满硬币的金库，前方是分别题为“储备证明通常包含”和“但无法证明”的清单 https://joviantan.com/zh/insights/binance-security-crisis-customer-trust https://joviantan.com/images/insights/binance-crisis-of-trust-2-zh.png 安全事件之后：币安能否把危机转化为加密行业的信任新标准？ 插图，标题为“遭黑客攻击！信任危机爆发？”。币安标志下方是标题和“这次事件揭示了什么？交易所的「信任标准」应该是什么？”，旁边是一把带币安标志的破裂挂锁、一个戴兜帽的人影和一幅 K 线图