A team of researchers has released an open-source tool meant to make EU AI Act compliance testing less chaotic.
The EU's AI Act requires every member state to set up an AI Regulatory Sandbox by August 2027: supervised environments where government regulators, technical experts, and the AI companies being assessed work together before a system ships. The researchers found the tools used for this kind of testing are fragmented, producing outputs that are hard to compare or reuse across cases. Their response is the AI Assessment Sandbox Configurator, a framework built around a catalogue of tests accessed through a shared plug-in interface, a common data model, and dashboards tailored to different reviewers. It has already been used in a live sandbox engagement, feeding into an official government Exit Report.
That timing matters. Twenty-seven countries now have under two years to stand up testing infrastructure that, right now, mostly doesn't exist in any standardized form. A shared, open-source toolkit could spare each country from building its own compliance-testing pipeline from scratch.
Still, this is one early pilot, not an adopted EU standard. Getting dozens of national regulators to converge on the same assessment software is a slower, messier process than writing the code.