Witnium logo
WITNIUM. Chain.

← Earlier briefings

Today in evidence · Sunday, 11 October 2026Published 12:50

Audit logs, product liability and the witness problem

This morning: marking AI content, who pays when ChatGPT goes wrong, and why the AI writing its own audit log is a conflict of interest.

Vera is an AI-generated presenter. Image and voice are made with AI; the script is taken word for word from the published briefing.
  1. 01

    Article 50 now enforceable for generative AI

    Providers must mark synthetic audio, image, video and text in machine-readable format. The rules apply even to systems released before 2 August 2026.

    Why it matters for evidence: The marking tells you it came from AI. Proving it hasn't changed since is a separate question.

    Source: complylayer.comOur explainer →

  2. 02

    US courts test whether ChatGPT is a product

    OpenAI argues ChatGPT is a service to avoid strict liability claims. The classification changes who bears the burden of evidence when AI causes harm.

    Why it matters for evidence: Under the EU's new product liability rules, software is a product. The side without records starts on the back foot.

    Source: reuters.comOur explainer →

  3. 03

    AI Act meets MDR for medical devices

    AI systems that are medical devices or safety components become high-risk under Article 6, triggering automatic logging, data governance and third-party assessment.

    Why it matters for evidence: Post-market surveillance requires audit logs that hold up. Both MDR and the AI Act expect them.

    Source: johner-institut.de (in German)Our explainer →

  4. 04

    The witness was the suspect

    An audit log written by the AI being audited creates a conflict of interest: a malfunctioning system could falsify its own trail. This undermines the log's evidentiary value.

    Source: daily.devOur explainer →

  5. 05

    Akamai's governance framework for AI agents

    The framework recommends capturing input prompts, execution plans, tool calls with results, data accessed and final outputs to ensure traceability for agentic AI.

    Source: akamai.comOur explainer →

  6. 06

    Electronic evidence transforms maritime investigations

    Shipboard technology reconstructs incidents and decisions. The broader challenge is preserving, authenticating and assessing electronic records for reliability in legal contexts.

    Source: ibj-online.comOur explainer →

Every story here comes down to the same question: can you show what happened, and when? That's what a digital witness is for. How it works

Want the whole week in five minutes? Get it every Friday

Seal pending