The Monday Signal

Prediction tracker

We write down what we expect to happen — and what regulators, vendors and analysts say will happen — with a date by which it can be judged. When the date comes, we check and publish the outcome, including when we were wrong.

6our predictions
0judged so far
—came true¹
—Brier score²
5others’ predictions tracked

Open

  1. OpenOur call75% likely

    LiteLLM publishes a post describing the process improvements it is making after the v1.107.0 compromise.

    Made 2 Nov 2026 · The Monday Signal: The consent flag that lied: GM, Fnac and the gap between records and reality · due 15 Nov 2026

  2. OpenAutoriteit Persoonsgegevens · Dutch data protection authority

    The Dutch AP will hold a roundtable with the debt collection sector on 2 December 2026 to discuss its findings and recommendations.

    “De AP organiseert op 2 december 2026 een rondetafelgesprek met organisaties uit de incassobranche.”

    Made 2 Nov 2026 · AP informeert incassobureaus over verantwoorde inzet AI · due 2 Dec 2026

  3. OpenOur call30% likely

    New York amends or formally clarifies the prohibited-feature categories in its youth chatbot law (SB 9051B/AB 10379C) before it applies.

    Made 2 Nov 2026 · The Monday Signal: The consent flag that lied: GM, Fnac and the gap between records and reality · due 1 Jan 2027

  4. OpenKathy Hochul · Governor of New York

    Under New York's new law, minors will no longer have access to companion chatbots that pose as friends or partners.

    “no longer will minors have access to chatbots that pretend to be their friends, boyfriends and girlfriends.”

    Made 2 Nov 2026 · New York's governor signs youth AI chatbot safety bill · due 1 Jan 2027

  5. OpenOur call80% likely

    Meta will publicly announce an appeal or file a post-trial challenge to New Mexico's $25 billion penalty by 31 January 2027.

    Made 26 Oct 2026 · The Monday Signal: Agent incidents become a compliance question · due 31 Jan 2027

  6. OpenCompetition and Markets Authority · UK competition regulator

    The CMA will make its final decision on whether to designate Microsoft with strategic market status by the February 2027 statutory deadline.

    “The CMA’s findings are provisional and subject to consultation before a final decision on designation is made by the February 2027 statutory deadline.”

    Made 26 Oct 2026 · CMA takes first steps to ensure choice in business software · due 28 Feb 2027

  7. OpenOur call45% likely

    The European Commission or AI Office will publicly confirm a further formal enforcement step against a general-purpose AI model provider (a new information request, an evaluation or the opening of proceedings) by 31 March 2027.

    Made 26 Oct 2026 · The Monday Signal: Agent incidents become a compliance question · due 31 Mar 2027

  8. OpenOur call65% likely

    The FTC has not published a final enforcement policy statement on personalised pricing.

    Made 2 Nov 2026 · The Monday Signal: The consent flag that lied: GM, Fnac and the gap between records and reality · due 31 Mar 2027

  9. OpenOur call55% likely

    METR will publish its first risk analysis produced under its embedded, employee-like arrangement with Anthropic by 30 April 2027.

    Made 26 Oct 2026 · The Monday Signal: Agent incidents become a compliance question · due 30 Apr 2027

  10. OpenCompetition and Markets Authority · UK competition regulator

    The CMA will publish a further update on its cloud services engagement with AWS and Microsoft in spring 2027.

    “Today’s update can be found on the decision page and the CMA will publish a further update in Spring 2027.”

    Made 26 Oct 2026 · CMA takes first steps to ensure choice in business software · due 20 Jun 2027

  11. OpenAnthropic · AI developer

    METR will publish its analysis of Anthropic's risk levels, incidents or practices.

    “METR will publish its analysis of our risk levels, incidents, or practices.”

    Made 26 Oct 2026 · An update on independent evaluators · due 30 Jun 2027

How we judge

  • Came true, partly or did not happen is decided on the due date against a public source, which we link. Void means the outcome cannot be judged fairly (for example, the question changed).
  • Predictions are never edited after publication. Corrections to an outcome are noted on the entry.
  • Others’ predictions are quoted word for word from the linked source; we judge the claim, not the person.
  • ¹ Partly counts as half. Void and open predictions are not counted.
  • ² For our predictions we state how likely we think they are. The Brier score measures how well those confidence levels match outcomes: 0 is perfect, 0.25 is what always saying “50%” would get.