What if the token says you own the security, but the law says the real record is somewhere else?
I ran into that question while reading Dusk’s latest article on SME tokenization.
The article gives a concrete Dutch example: transfers of BV shares require a notarial deed.
That creates a question I hadn't really considered. If the security is represented on-chain, but a legally required process still sits outside the chain, what exactly is the token representing?
I’d been thinking about tokenized ownership mostly as a question of putting the asset on-chain. But the harder part may be keeping that digital ownership state aligned with whatever record the jurisdiction actually recognizes.
If those two states can ever disagree, tokenization hasn't completely removed reconciliation. It has created a new coordination problem between the digital and legal sides.
So when the on-chain ownership state and the legally authoritative record disagree, which one does Dusk treat as the source of truth?
After the warning, a Dusk provisioner can have 10% of its stake moved into Rewards, but the tokens aren't burned.
That was the part I didn't expect.
Dusk's finalized soft-slashing mechanism escalates with consecutive faults. N faults means N × 10% of the stake is moved into that same node's Rewards balance, while the provisioner is excluded from consensus for N epochs.
So the penalty isn't simply “your tokens disappear.”
The stake stays with the same provisioner. What changes is how much of it remains active for consensus.
There's another detail I found even more interesting. The fault count doesn't reset just because the suspension ends. Dusk says the warning and fault count reset when the provisioner actually earns a reward by producing a block or successfully voting.
So waiting isn't what restores the record. Participating successfully is.
The active-stake reduction can also continue toward the network's 1,000 DUSK minimum.
I started thinking about soft slashing differently after reading that. It's less about taking someone's tokens away and more about progressively reducing the active weight and eligibility of a provisioner that keeps failing.
Does that make recovery from repeated faults intentionally harder than simply waiting out a suspension?
I thought a fast DuskEVM transaction was basically a settled transaction.
Then I found a warning in Dusk’s docs that made me rethink that assumption.
DuskEVM separates transaction inclusion from settlement.
A transaction can be included in an L2 block quickly, but that doesn’t mean the resulting state has already settled back to the Dusk L1. The two stages are connected through batching, state commitments and fault proofs.
The detail I found most interesting is that @Dusk explicitly tells applications moving value between DuskEVM and the Dusk L1 NOT to infer finality simply from elapsed time.
That sounds obvious after reading it, but it’s actually an important design distinction.
“Confirmed quickly” and “safe to treat as settled” are not necessarily the same thing.
For an application moving real value, using a timer as a shortcut could mean acting on inclusion while the cross-layer settlement process is still incomplete.
So I’m left with one question:
What exact protocol status should an application treat as authoritative before releasing value across the DuskEVM ↔ Dusk L1 boundary?