Can I describe a confidential deal?
+
A description you type is read once and not stored, unless you save it as a project. To read it, the text is sent to a language model provider, which may process it outside the EU. So describe the deal without names: the products, the countries and the structure are what the search uses. During the test period, do not paste client-confidential material.
What does it cover?
+
Mergers notified to the European Commission under the EU Merger Regulation, from 1990 to today: more than 10,300 cases in the Commission's register. More than 9,900 of them have a published decision that has been read and coded, with more than 45,000 markets and 25,000 theories of harm. The others have no published decision, or one that could not be read. The Cases page in the app shows the status of every case.
Does it cover cartels or abuse of dominance?
+
Not yet. Today it covers merger control only: cases notified to the European Commission under the EU Merger Regulation. The Commission's cartel and abuse of dominance decisions are coming soon. Decisions of national competition authorities are not in it.
Where does the data come from?
+
From the European Commission's public case register and the decisions published there. We add no commentary of our own. Facts taken from the register, such as dates, the phase and the deadlines, are stored and labelled apart from values read from the decision text.
How current is it?
+
The Commission's case register is checked daily. A new decision is picked up and read shortly after the Commission publishes it, and cases still under review are listed with the Commission's provisional deadline.
Who reads the decisions, a person or a model?
+
A language model reads each decision into the scheme, part by part. Automatic checks flag a reading that looks incomplete, and a flag is a prompt for a person to look at the case. The scheme itself, the variables and the values they can take, is set by a competition lawyer, who checks readings against the decisions.
How do I know an answer is right?
+
You check it. Every claim points to the paragraph of the decision it came from, and that paragraph opens next to the answer. An answer can be wrong, so the reference is the part to rely on, not the summary.
Is it a chatbot?
+
No. The record is the product: every case coded into the same variables, each value with its paragraph. The Analyst does answer a question in words, but it builds the answer from the coded data and lists the cases it used underneath.
Does it write my memo or report?
+
No. It gives findings with their references: cases, tables and passages. You copy what you need into your own work and draw the conclusion yourself.
Can I use the data in my own AI assistant?
+
That is being tested. The coded decisions can be reached over MCP, the open standard for connecting an AI assistant to a source. It is not open to everyone yet. Tell us which assistant you use when you ask for access.
Is it legal advice?
+
No. It is a research tool. It does not give legal advice and it is not a substitute for reading the decision or for professional judgement.
How do I get access?
+
Access is by invitation during the test period. Fill in the request form and tell us briefly what you work on.
Who is behind it?
+
Two people. Jasper Sluijs, associate professor of competition law and regulation at Utrecht University, decides what is coded and how. Maurits Fornier, a former litigator who builds legal tools, builds the pipeline and the search. Seek is not affiliated with the European Commission.
Why the name Seek?
+
It is a play on SIEC, the significant impediment to effective competition: the test a merger is assessed against under the EU Merger Regulation. It is also what you do here: seek the case, the market or the paragraph you need.