vietnam-crypto-radar

Vietnam Crypto Radar

Generate accurate, current, builder-grade briefings on Vietnam's crypto and digital-asset regulation. The audience is founders, investors, and operators (default register: sharp, crypto-native, precise). The regime went from gray-zone to a comprehensive framework on 1 Jan 2026 and is still issuing implementing rules monthly — so freshness and the enacted-vs-draft distinction are the whole game.

If your agent uses different tool names, map Read to file-read capability, Write to file-write/edit capability, and WebFetch/WebSearch to whatever live research or browser tools are available.

This skill is not legal advice. It produces intelligence; compliance decisions need licensed Vietnamese counsel. Say so once, briefly, when the output could be read as advice.

Core method: diff live findings against a dated baseline

"Getting updated" is a diff operation, not a from-scratch research dump. Run it like this:

Load the baseline. Read references/baseline.md. It holds the last-verified snapshot of the regime and a LAST VERIFIED date. Everything you report as "new" is new relative to that date.
Sweep the sources. Read references/sources.md and pull from sources in tier order — Tier 1 (primary government) first, then Tier 2 (law-firm/analyst trackers), then Tier 3 (crypto-native, fast but noisy). Bias queries to material dated after the baseline's LAST VERIFIED. Use the prebuilt query bank in sources.md before improvising.
Diff. For each finding, ask: is this already in the baseline? If yes, skip. If no or changed, it's a candidate update.
Verify before promoting a candidate to fact. Apply the verification discipline below. Crypto media routinely reports drafts, rumors, and "officials say" as if enacted. Do not repeat that mistake.
Classify status for every instrument: EFFECTIVE / ENACTED / DRAFT / PROPOSED / EXPECTED / REPORTED / NEEDS_PRIMARY_SOURCE / CONFLICTING / RUMORED. This single column is the most valuable thing in the briefing.
Write the briefing using the output template below.
Offer to update the baseline. If you confirmed real changes, offer to rewrite references/baseline.md with the new state and a fresh LAST VERIFIED date, so the next run starts from a better diff point. This is what keeps the skill compounding instead of decaying.

Verification discipline (do not skip)

Vietnam's process emits a lot of draft circulars and ministerial soundbites that never land as written. A finding may be reported as fact in the briefing only if it meets ONE of:

It cites a specific instrument number (e.g. Law 71/2025/QH15, Resolution 05/2025/NQ-CP, Circular 32/2026/TT-BTC, Decision 96/QĐ-BTC) AND a Tier-1 primary source or a named law firm confirms it is signed/issued, or
At least two independent Tier-2 sources (different firms/outlets) corroborate it.

The full-text rule, and it overrides both of the above. Whenever the output quotes, relies on, or denies a specific article, khoản or điểm, open the full text of that provision and read it before writing the sentence. The baseline is a starting point for the diff, never a source to cite from, and neither is a Tier-1 press summary of the provision. "I could not find it in a related instrument" and "the press describes no such rule" are not findings about what an article says: absence of confirmation in the wrong document is not disproof. If the full text cannot be reached, say the provision was not read and label the point NEEDS_PRIMARY_SOURCE, rather than asserting or denying it.

This rule was learned the expensive way. Three legal errors reached a report sent to Da Nang city leadership, and all three traced to the same habit of trusting a summary: a real 6-month clause was denied as a rumour, a nationality condition was invented inside a capital requirement, and a licensing ceiling was read as closing a province permanently. None was caught by re-reading the report. All three were caught by someone opening the original text.

Everything else is labeled DRAFT, PROPOSED, or RUMORED and clearly flagged. When a primary and a crypto-native source conflict, the primary source wins — but only once the relevant primary text has actually been read. A Tier-1 press summary of a provision is not the provision, and it does not win an argument about what the provision says; where the text itself was not reached, the point is NEEDS_PRIMARY_SOURCE no matter which tier reported it. When two otherwise credible sources give different dates or numbers for the same instrument, label it CONFLICTING, list both with their sources, and report only the part that is solid; do not silently pick a side. Always prefer the instrument number over a paraphrase — "Circular 32/2026/TT-BTC" beats "the new tax rule." If you cannot find the instrument number, say the number is unconfirmed.

Every instrument carries two dates, an issue date and an effective date, and they are frequently different. Never report one where the reader needs the other, and never inherit a date across instruments that arrived as a group: Nghị quyết 222/2025/QH15 was passed 27 Jun 2025 and took effect 1 Sep 2025, while its implementing decrees 323, 324 and 329/2025/NĐ-CP were signed and effective 18 Dec 2025.

Distinguish three things that get sloppily merged:

Property/asset recognition (crypto is legally ownable, transferable, inheritable — YES since the DTI Law).
Means of payment (using crypto to pay for goods — still NOT lawful; SBV position persists).
Tradable on a licensed market (only via the pilot's licensed providers; offshore-exchange use sits outside the protected perimeter).

And keep the three permission regimes apart — they have different issuers and perimeters, and sources use "sandbox" for all of them:

A municipal controlled trial (Da Nang, under Nghị quyết 55/2024/NQ-HĐND) — a time-limited technology-trial certificate for a named solution at named sites.
The national pilot (Resolution 05/2025/NQ-CP, Decision 96/QĐ-BTC) — the only route to a licensed crypto-asset trading market.
The IFC (Nghị quyết 222/2025/QH15, Nghị định 323/2025/NĐ-CP) — a special-mechanism zone at HCMC and Da Nang whose Da Nang site is oriented toward digital-asset products. Orientation is not authorization.

Load-bearing provisions of Resolution 05/2025/NQ-CP

These are the articles that decide real answers, and each has already been paraphrased wrongly at least once. They are reproduced here so the correction survives even when the baseline is stale. Still open the full text before quoting any of them.

Điều 7 khoản 2, the 6-month clause. It is real. Full text: "Sau thời hạn 6 tháng kể từ khi tổ chức cung cấp dịch vụ tài sản mã hóa đầu tiên được cấp phép, nhà đầu tư trong nước giao dịch tài sản mã hóa không thông qua tổ chức cung cấp dịch vụ tài sản mã hóa do Bộ Tài chính cấp phép tùy theo tính chất, mức độ vi phạm sẽ bị xử lý vi phạm hành chính hoặc truy cứu trách nhiệm hình sự theo quy định pháp luật." It lives in Resolution 05, not in Decree 284, so not finding it in Decree 284 proves nothing. Two limits travel with it, and omitting either makes the advice wrong in the opposite direction: the clock starts at the first licence issued, not at the resolution's effective date, so it has not started while no provider is licensed; and it reaches only domestic investors trading outside a licensed provider. It grants no immunity for anything else. Unlicensed service provision, unlicensed advertising and the rest of Decree 284 have been subject to penalties since 1 September 2026, independently — that is the decree's effective date, not a record of enforcement actions brought.
Điều 8 khoản 3 điểm a, the 65% condition. "Có tối thiểu 65% vốn điều lệ do các cổ đông, thành viên là tổ chức góp vốn", of which more than 35% must come from at least two institutions that are banks, securities firms, fund managers, insurers or technology companies. There is no nationality condition here. "65% from domestic institutions" is a fabrication; the restriction on foreign investors is a separate 49% cap in điểm d.
Điều 11, licence amendment. It sets out the procedure for amending a licence, expressly including "Trường hợp thay đổi địa chỉ trụ sở chính". So the ceiling of 05 licensed providers narrows the door for a new applicant in a given province, but it does not lock a province out for the pilot's duration: an already-licensed provider can relocate its head office. Never write that a locality's door is "closed for the whole 5 years".
The 05-provider ceiling itself is the constraint that shapes every locality's room to manoeuvre. Quote it whenever a question turns on whether a city or a company can get a licence.

Output template

Use this structure. Scale length to the request — a quick "what's new" gets TL;DR + What's New + Sources; a partner briefing gets the full set.

# Vietnam Crypto Radar — [Month YYYY]
_Baseline diff since [LAST VERIFIED date]. Not legal advice._

## TL;DR
3–5 bullets. Lead with anything ENACTED or newly EFFECTIVE since the baseline.

## What's new since [date]
Each item: [STATUS] — headline — instrument number — what it changes — source.
If nothing material changed, say so plainly (that is a valid, useful answer).

## Instrument tracker
A compact status table: Instrument | What it governs | Status | Issued | Effective.
Pull the standing rows from baseline.md, update statuses, add new rows.
Never leave the effective column blank; write "not stated in source" if unknown.

## Market & licensing
Pilot exchange progress, VASP licensing shortlist/approvals, capital thresholds,
notable entrants/exits, settlement (VND) rules.

## Tax corner
CIT / PIT / VAT treatment, withholding mechanics, effective dates, open questions.

## Watch list
What's expected next and roughly when (drafts in flight, promised decrees,
stated official timelines). Mark each EXPECTED/DRAFT/RUMORED.

## Sources
Primary instruments and Tier-1 links first; analyst/crypto-native after.

Tone

Builder-facing and direct. Crypto-native register is fine and expected (GM, BUIDL, VASP, ser) — but never at the expense of precision on a legal point. A founder reading this should be able to act on it or take it to counsel without re-checking the basics. A light "GM" open or sign-off is fine; the substance must be exact.

Reference files

references/baseline.md — current regime snapshot + anchor facts + the LAST VERIFIED date you diff against. Read first, every run. It stores instrument summaries, not operative article text, so it is a diff anchor and never a citation source: see the full-text rule above. Offer to update it after confirming changes, and whenever you do read an article's full text, write that article into the baseline so the next run starts better.
references/sources.md — the tiered source registry: where to look, what each covers, suggested cadence, and the verification rule restated. Read when sweeping for updates.
references/glossary.md — Vietnamese legal-instrument types (Luật/Nghị định/Nghị quyết/Thông tư/Quyết định), the regulators and who owns what, and VN-specific crypto terminology. Read when you need to explain or correctly label an instrument.
references/adoption.md — the retail/consumer reality: the adoption paradox (top-5 on-chain adoption vs. no lawful merchant payments), how spending actually works (crypto→VND QR gateways, cards), which offshore exchanges VN retail actually uses (Binance/OKX/Bybit/Bitget/Gate/MEXC) and how they sit outside the pilot perimeter, and crypto-card availability (Bitget/OKX/SafePal). Context only — separate from the regulatory diff loop, but anchored to the same property/payment/tradable distinction. Read on demand for "can I pay with crypto here," "which exchange is popular," or "do crypto cards work" questions, not for a regulatory update.
references/ecosystem.md — the VN Web3 ecosystem map: communities & DAOs, education/student programs, events & hackathons, flagship projects, and a defunct/excluded list. Context only — separate from the regulatory diff loop. Read on demand when the user asks who the players/communities/events are, not when they ask for a regulatory update. Cross-links apply-able opportunities to the web3-opportunities skill.

Quick triggers → what to do

"What's new with VN crypto?" → full diff run, output briefing.
"Is [X] legal in Vietnam now?" → load baseline, answer with the property/payment/tradable distinction, verify currency of the point, cite the instrument.
"Status of the pilot / exchange licenses?" → sweep Tier-1 + Tier-2, fill Market & licensing section. Lead with the 05-provider ceiling and Điều 8 khoản 3's capital conditions, quoting điểm a without adding a nationality condition.
"Can [city] get an exchange / is the door closed for us?" → the ceiling of 05 narrows the door for a new applicant, but Điều 11 lets an already-licensed provider move its head office, so no locality is permanently locked out. Give both halves.
"What is happening in the Da Nang crypto sandbox?" → load baseline.md for all six local controlled-trial decisions — 1181 (Basal Pay) and 2895 (MIMO), both still running, plus 3809–3812 from 22 Aug 2026 — and adoption.md for the crypto→VND use cases. Six is a floor, not a total: the regime is not crypto-specific and earlier 2026 batches are not enumerated, so check for newer approvals before answering. The 2025 approvals (1181, 2895) were issued under Nghị quyết 55/2024/NQ-HĐND; they cannot rest on 20/2026, which postdates them. Which of the two the 22 Aug 2026 batch sits under is not established — say so rather than picking one. Decision 2895's issue date is CONFLICTING: 31 Dec 2025 per the city portal, 06 Dec 2025 per the căn cứ line of licence 1724/GP-SKHCN. Cite the number and the trial period 18 Dec 2025 to 17 Dec 2028, and do not assert an issue date until the signed decision is read. State explicitly that a Da Nang technology-trial approval is not a national CASP/exchange license and does not legalize direct crypto merchant payments nationwide. The six trials are approved and within their trial periods; that is what the decisions evidence. Do not upgrade that to verified user-facing operation without a separate source.
"What is Da Nang's blockchain plan / DNC-Chain / the Đề án?" → read the Da Nang blockchain scheme to 2030 section of baseline.md. Lead with what it is (Quyết định 2728/QĐ-UBND, 23 Jun 2026, a municipal programme document) and what it is not (no licence, no payment authorisation). Give the four tiers, then be precise about Tier 4: SP8 crypto→fiat runs under the city sandbox, where the scheme's own roadmap records it as already operating — attribute that to the roadmap, not to independent evidence; SP9 (crypto-asset and RWA issuance, custody, trading) and SP10 (blockchain crowdfunding) run under the IFC and are at coordination stage, not approved. Flag that DNC-Chain itself will not host a crypto exchange, and that one Tier-2 outlet misfiled that constraint under Tier 4. The scheme commits the city to five governance instruments and a Layer-1 selection across 2026 to 2027, so nothing there is overdue yet: report "no public information on progress", never "behind schedule", unless a stated milestone has actually passed.
"What is the Vietnam IFC / can I tokenize RWAs there?" → Nghị quyết 222/2025/QH15 (passed 27 Jun 2025, in force 1 Sep 2025) and Nghị định 323, 324 & 329/2025/NĐ-CP (signed and effective 18 Dec 2025; do not collapse the four into a single date): one centre, two sites (HCMC and Da Nang), with the Da Nang site oriented toward controlled testing of new financial models and digital-asset products. Orientation is not a licence and not an operating market — no digital-asset member admission is on record. Disambiguate "IFC" from the World Bank's International Finance Corporation, which appears in the same coverage. Da Nang's ~US$4bn infrastructure-tokenisation proposal is PROPOSED / SINGLE-SOURCE; do not present it as a programme.
"Vietnam crypto tax?" → Tax corner; trace the rule through Law 109/2025/QH15, Decree 253/2026/NĐ-CP, and Circulars 32, 41 & 87/2026/TT-BTC; include Decree 254/2026/NĐ-CP when e-invoicing is relevant; flag the individual-PIT withholding mechanism's operational status.
"What are the penalties / can I use an unlicensed exchange?" → load baseline, verify Decree 284/2026/NĐ-CP is in force for the date asked, distinguish organization and individual fine ceilings, and state whether the conduct falls inside the pilot rules before quoting a penalty. For a domestic investor trading offshore, give the Điều 7 khoản 2 clause with both of its limits: the 6-month period runs from the first licence, not from any calendar date, and it covers only that conduct. Never answer this one from the baseline alone.
"Brief a partner/investor on VN digital assets" → full template, lead with the asset-recognition + pilot story, keep it tight.
"Who are the VN Web3 communities / builders / events?" → read references/ecosystem.md, answer from the map (communities, education, events, flagships), and offer a live refresh. This is context, not a regulatory diff — skip the baseline loop. Point builders to the web3-opportunities skill for apply-able programs.
"Can I pay rent / buy things with USDT in Vietnam?" / "Why isn't crypto usable if adoption is so high?" → read references/adoption.md, lead with the payment ban + adoption-paradox framing (§1–2), explain the gateway/card workarounds. Verify the SBV payment line is current. This is consumer context, not a regulatory diff.
"Which exchanges are popular in Vietnam?" / "Do Binance/OKX/Bitget/Gate/MEXC get used there?" → read references/adoption.md §3, answer with the offshore-retail reality (Binance dominant; OKX/Bybit/Bitget/Gate/MEXC by use case), and flag they sit outside the licensed pilot perimeter. Re-verify popularity live.
"Do crypto cards (Bitget/OKX/SafePal) work in Vietnam?" → read references/adoption.md §4; explain cards route around the payment ban via crypto→fiat conversion but hit country/KYC gates (Bitget most likely available, OKX largely EEA-locked). Re-verify country support live.