Estonia tests AI for finding errors in draft laws
A beta prototype analyses legislative texts to flag broken references, contradictions and impossible dates before a bill is approved.
On January 14, 2026, a tool was unveiled in Estonia for a very specific AI use: reviewing draft laws before they are approved. The debate had been opened by a drafting error in an amendment to gambling tax law that inadvertently exempted online casinos during 2026, according to Estonian Public Broadcasting service ERR, and that reopened a simple question: can a machine help find mistakes that a human review chain did not catch?
What is being tested is not AI that writes or decides laws. Luukas Kristjan Ilves, a former digital-development official at the Ministry of Economic Affairs and Communications, introduced a beta prototype called Apsakaleidja, or mistake finder. Its role is to read bills under consideration on the Riigikogu website, Estonia’s parliament, and flag elements that deserve a second review.
What the prototype looks for
According to Ilves’s description reported by ERR, Apsakaleidja can flag inconsistent use of defined terms, broken references to other sections, contradictions between legal text and explanatory memoranda, arithmetic errors and impossible dates. These are issues that can, in principle, be framed as consistency checks across public documents.
That gives the tool a clear boundary. It does not determine whether a policy is fair, constitutional or wise. It does not decide which economic consequences a country should accept. It attempts to draw attention to anomalies that may be hidden among many edits, cross-references and deadlines.
Ilves himself described the system as a beta that can be improved. That detail matters: detecting an inconsistency is not the same as proving a legal mistake. An alert still needs review by people who understand the legislative context, the intent of the text and its relationship with existing law.
An extra check, not an authority
Estonia’s public debate has maintained that boundary. In WIRED’s report on the project, Prime Minister Kristen Michal stressed that AI does not replace democratic institutions or the will of voters. If AI identifies a mistake, responsibility for correcting it remains with parliament, courts or public administration.
The idea matches another strand of Estonian government work. In June, its Eesti.ai advisory board agreed to move forward with digital identities for AI agents, with defined limits and actions that are verifiable and auditable. The principle is not to give a system a blank check, but to know who is acting, under which authorization and who is ultimately accountable.
Applied to legislative review, that suggests a useful rule: a tool can assist the process, but it should leave a record of what it analysed, what it flagged and how each alert was resolved. Traceability makes it possible to discuss system errors without diluting institutional responsibility.
The challenge is not only technical
Estonia’s parliament has acknowledged remaining shortcomings in the quality of some draft laws, including insufficient constitutional analysis. A consistency tool could help free time for those more demanding reviews, but it cannot replace them.
It can also produce false positives, fail to understand a deliberate exception or miss a deeper problem written in perfectly consistent language. Its best use is therefore likely to be an additional layer of quality control, embedded in a process with experts and a clear human decision.
The point of the Estonian case is not to promise an automated legislator. It is to show a concrete administrative task where AI may help: finding details that deserve another look, earlier. In lawmaking, spotting a signal is valuable; deciding what to do with it remains a democratic responsibility.
Sources
Sources for this piece
This piece draws on 4 primary source(s), gathered during reporting.
This article was produced with artificial intelligence under human editorial oversight.
