A single wrong word in a law cost Estonia roughly $27 million a year in tax revenue. Parliament passed an update to the gambling tax act in December 2025, intending to lower the rate slightly on remote gambling. But the wording covered only "skill games," not games of chance, which meant online casinos were legally outside the tax net for the entire year. The Finance Ministry apologized. A parliamentary advisor was fired. The casinos, to their credit, kept paying voluntarily anyway, labeling the payments as "donations" to the state treasury because that was the only legal mechanism available.
What made the story useful rather than just embarrassing: when someone ran the flawed bill through two widely available AI tools, both spotted the problem within seconds. The bill had been read by lawyers at the Finance Ministry, committee members, and the president. None of them caught it. The AI did.
A former government official built a rough prototype called the Fuckup Finder in hours. It pulls draft bills directly from parliament's website and flags broken references, contradictory wording, arithmetic errors, and impossible dates. Of the 112 bills currently listed on the site, 102 are rated high risk. He demonstrated it on national television.
This is the pattern worth paying attention to. AI is not replacing the judgment call, the policy debate, or the political decision. It is doing the mechanical consistency check that humans are bad at, especially when tired, rushed, or under political pressure. The gambling tax law was passed quickly under difficult political circumstances. A lawyer at the ministry had reportedly warned before the vote that the bill needed more time. Nobody listened. The AI would have flagged the problem before any of that happened.
Estonia's government turned the incident into a policy direction. The prime minister launched the Eesti.ai program in January 2026, targeting a 25% increase in GDP within five years and a doubling of labor productivity by 2035. The numbers are ambitious; the program is still in early stages and Estonia's public finances are under pressure, with a government deficit forecast at 4.4% of GDP in 2026. But the direction is clear and the structure is serious: the advisory board includes the founder of Bolt and meets quarterly with the prime minister in attendance.
A bill currently working through parliament would give state and local government the legal right to automate administrative decisions using AI. The proposed framework draws a practical line. Where the law determines the outcome from verifiable facts, automation is appropriate: you qualify for a benefit, you get told so, without having to fill in a form. Where a decision requires weighing competing interests or judging specific circumstances, a human stays involved from the start. Citizens can invoke the right to a human review at any point, and every automated decision must leave an audit trail.
In June 2026, Estonia went a step further. The prime minister approved a proposal to issue official digital IDs to AI agents, making Estonia the first country to do this. The idea is straightforward: when an AI acts on your behalf, say filing a tax declaration, booking a service, or interacting with a government portal, it should carry a verifiable identity that defines exactly what it is allowed to do and who is responsible if something goes wrong. Today, most AI tools borrow their user's full credentials, which creates obvious problems. An AI ID code would scope the permissions and create an auditable trail.
For business operators outside Estonia, this matters in two ways. First, the mechanical lesson: an AI reading your own contracts, proposals, or internal policies for consistency errors is already possible today with standard tools and costs almost nothing to try. Second, the broader pattern: governments are beginning to define rules for AI acting on behalf of people and companies. Estonia is moving first. Whatever it builds will likely influence how the EU handles the same question, just as Estonia's earlier digital identity work shaped EU-wide standards. The rules your AI tools will eventually operate under are being written now, and this is where they are being written.