Labels, logs and a week to patch: four rules meeting reality
The AI Act's transparency rules go live, a health insurer's algorithm faces court, and attackers exploit most vulnerabilities within seven days.
- 01
Commission says deepfakes need labels
The European Commission announces that deepfakes must carry clear labels to help people recognise AI-generated or manipulated content. This implements Article 50(4), which has applied since 2 August 2026.
Why it matters for evidence: If you make synthetic media, you have to show what was AI-generated and what wasn't, on the day someone asks.
Source: linkedin.com
- 02
Health insurer's AI faces allegations
Private insurers allegedly used the nH Predict algorithm to cut off post-acute care coverage for elderly Medicare Advantage patients based on false diagnoses. The case raises questions about whether insurers can prove what their AI systems decided.
Why it matters for evidence: When an AI cuts off someone's care, you have to show what it decided and why, not what you think it probably did.
Source: airisk.mit.edu
- 03
AI-generated ads need visible labels
From 2 August 2026, all newly published AI-generated advertising content meeting the requirements must be visibly labelled as AI-generated. This implements Article 50's transparency obligations for generative AI in advertising.
Why it matters for evidence: The label proves you disclosed it. Without one, the question is whether you can show the content wasn't AI-generated.
Source: beyond.global
- 04
Most vulnerabilities exploited within a week
Attackers exploit 71% of high-severity vulnerabilities within a week of disclosure. That leaves minimal time to patch systems before exploitation occurs.
Source: Help Net Security
- 05
VAST Data logs every AI agent call
VAST Data describes how production AI agents require transactional logging of every tool call, prompt, token and decision to create an audit trail that persists after sandboxes are destroyed.
Source: linkedin.com
- 06
Voice AI for debt collection goes live
MEGA.AI offers deterministic voice AI for lenders and collection agencies covering the entire journey from first missed payment to agency placement. The system raises questions under high-risk AI rules for credit assessment and automated decision-making.
Source: mega.ai