Countries score below 50 out of 100 on the Corruption Perceptions Index 2025. The global average fell to 42, the lowest in a decade.
Government records nobody can rewrite without everyone noticing.
HardState hands every citizen a signed receipt for each step of a procedure and anchors a 32-byte fingerprint of the public record to Bitcoin every ten minutes. Edits, deletions and backdating become provable by anyone, without trusting the office that keeps the records.
- 36 bytes
- per daily anchor in Bitcoin
- 960 bytes
- to prove one record among a billion
- Zero
- personal data on-chain
OP_RETURN 485353 01 93796826…✓ anchored — provable by anyoneEvery bribe at a government window leaves a record that the same office can quietly edit.
Countries measure corruption through surveys and indexes. What no survey can measure is how often the paper trail itself was changed, because the only copy belongs to the office being audited.
Paid in bribes worldwide every year, about 2% of global GDP, according to the IMF.
Public service users in Asia paid a bribe in the past year: 836 million people across 17 countries.
Bribes in public procedures cost MXN 17,707 million in 2025, up from 11,910 million in 2023. Business licenses are the second most corrupt procedure in the country.
When the record keeper is also the only witness, nobody can prove what really happened.
Sources: Transparency International CPI 2025; IMF, Corruption: Costs and Mitigating Strategies (2016); Global Corruption Barometer Asia 2020; INEGI ENCIG 2023 and 2025, constant 2018 pesos.
How a record becomes impossible to rewrite quietly
Five steps, all of them standard cryptography. Personal data never leaves the government database.
The office signs each step of the procedure and hands the citizen a HardState Receipt with its record number.
98c13d08dbb12435ceae9cd2d73370c7d434c04ecafe2627a6cd7a6310974e6d375dcfa2c15eea36e5178c912f0ec1a6187ed28b9647bd691c9f28d1c8cbf958b54881prevc4390598c13dprevb548814443dbprev98c13d0f3c93prev4443dbMerkle root
93796826e3229a40…OP_RETURN485353 01 93796826e3229a4024cb8b38…36 bytes- Fingerprint matches
- Office signature is valid
- Merkle path checks out
- Root found in block 921,384
Change one letter, change the whole fingerprint
Type anything. Your browser computes its SHA-256 fingerprint. Edit a single character and about half of the fingerprint flips, which is why nobody can tweak a record quietly.
With a salt, even a guessable text gets a fingerprint nobody can reverse. HardState salts every record.
The lab
This runs the real engine in your browser: SHA-256, Ed25519 signatures and RFC 9162 Merkle proofs. Only the Bitcoin network is simulated. Play the corrupt insider, then run the audit.
Office log
What the government database holdsMerkle tree
anchorednot anchored yetBitcoin (simulated)
Play the insider
Pick a record in the log, then tamper with it. Every attack covers its tracks: it renumbers, re-chains, re-signs and re-anchors.
Audit result
Original design
Root in Bitcoin, transaction id kept in the government database
Run the audit to compare both designs against the same database.
HardState
Same anchor, plus record numbers, a hash chain, a public log audit and citizen receipts
Run the audit to compare both designs against the same database.
Click any record or leaf to verify it step by step.
What the simulation shows
We attacked two implementations of the same office, 600 times per attack, with about 240 records a day. The insider holds the signing key and covers their tracks. One in four citizens checks their receipt. Each attack below is run against both systems: the government database as it works today, and HardState.
Same attack, two systems
Eight cover-up attacks, 600 simulations each — share caught when one in four citizens checks their receipt.
The database an office keeps can hide almost every cover-up. HardState catches them because each one breaks a citizen’s receipt or leaves a gap anyone can check. In the no-attack control, both raise 0% false alarms.
Detection rises with each citizen who checks a receipt
Edit a record before the next anchorDelete a record and renumber the restInsert a backdated recordDetection follows 1 − (1 − p)<sup>k</sup>, where k is the number of receipts an attack invalidates. Covering one deletion breaks every receipt after it, so cover-ups get harder to hide the more they touch.
Show the numbers as a table
| 0% | 5% | 10% | 25% | 50% | 75% | 100% | |
|---|---|---|---|---|---|---|---|
| Edit a record before the next anchor | 0% | 23% | 42% | 71% | 94% | 98% | 100% |
| Delete a record and renumber the rest | 0% | 22% | 42% | 68% | 86% | 98% | 100% |
| Insert a backdated record | 0% | 25% | 43% | 71% | 89% | 97% | 100% |
Method, code and raw numbers are in the repository: sim/run.mjs and sim/RESULTS.md. Deleting the last record of a batch can leave no gap in the numbering, which is why the receipt matters even for simple deletions.
Anchoring a whole country costs a few hundred dollars a year
We simulated 1,000 years of Bitcoin fees (calm, busy and spike periods) and a volatile Bitcoin price for five anchoring strategies.
Annual anchoring cost, USD, median and 5th to 95th percentile
Try your own numbers
The real cost is integration and training, not anchoring. That is why we start with one procedure in one city.
Why a public blockchain at all
Many countries already offer legal timestamping, such as qualified timestamps under eIDAS in the EU or NOM-151 certificates in Mexico. HardState keeps them and adds a witness nobody can lean on.
| Question | Government database | Database + legal timestamp | HardState: legal timestamp + Bitcoin + receipts |
|---|---|---|---|
| Who can rewrite history quietly | Any administrator | The office, with the provider’s cooperation | Nobody |
| Who can verify | Whoever the office allows | Courts and parties to a dispute | Anyone with a browser |
| Detects deleted records | No | No | Yes, with record numbers and receipts |
| Survives a change of administration | Depends on the next one | While the provider exists | Yes, Bitcoin has run since 2009 |
| Legal standing | Internal record | Recognized evidence | Recognized evidence, plus public proof |
We anchor 36 bytes per transaction. No token, no cryptocurrency for citizens, no personal data on-chain.
What HardState does not do
- It can’t tell if a record was true when it was entered.
- A fake inspection signed on day one stays fake. We reduce the risk with two signatures per step (officer and citizen) and fingerprints of the evidence, and leave the rest to auditors.
- It can’t force an office to issue receipts.
- Adoption needs a mandate. We propose starting where a mayor wants to prove clean licensing.
- It depends on someone checking.
- Anyone can audit the whole log in the browser, and a gap in the numbering shows up on its own. Citizens check by scanning a code; the evidence above shows how fast detection grows.
- Activity volume is visible.
- One national root per day hides how many records each office produced.
Built on a decade of precedent
Anchoring government records works. What earlier systems lacked was a receipt in the citizen’s hands.
- Estonia, since 2012KSI hash chains protect health, court and business registries against insiders. Citizens can’t check them on their own.
- Georgia, since 2017Land titles anchored to Bitcoin through a private chain. It worked because the registry was already digital and trusted.
- Argentina, since 2017The Official Gazette is stamped daily on the federal blockchain BFA, free of charge.
- Honduras, 2015A Bitcoin land registry announced with a startup stalled within months. The lesson: politics first, technology second.
From a working product to a public good
HardState is open source and designed to be run by many hands: offices write, universities and civil society audit, anyone verifies.
- 1
October 2026
Core live on Bitcoin mainnet with passkey signing; open simulation and this demo.
- 2
Pilot in Mexico, 3 months
Business licenses in one city, the country’s second most corrupt procedure. Receipts by QR, a daily anchor, a local timestamping provider, and a university auditing the public log.
- 3
National scale
Permits, inspections and public purchases. One national root per day, express anchors on demand, adaptive fees when the network spikes.
- 4
Open standard
Registered as a Digital Public Good so any country can plug in its own legal timestamping authority.
A public-good model: citizens verify free, forever; small institutions run free up to ~10,000 records a month; larger institutions pay ~$99–199/month, which only widens the free tier. Adoption starts with institutions that want to prove they are clean.
Serves UN Sustainable Development Goal 16: targets 16.5 (reduce bribery), 16.6 (transparent institutions) and 16.10 (public access to information).