Logs, deepfakes and *a systemic pattern* of misleading
From a US court ruling on Meta to machinery traceability and AI audit trails, today's news centres on proving what systems did and when.
- 01
Court finds systemic misleading in Meta's business model
A court in New Mexico ruled that Meta engaged in a systemic pattern of misleading practices at the heart of Facebook's business model. The case turned on whether Meta could demonstrate transparency and accountability for how its platform operates.
Source: heise online (in German)
- 02
Tracker maps when AI rules hit financial services
Meilynx published a detailed tracker showing when AI regulations apply in finance. Credit scoring and life insurance pricing are high-risk uses, with obligations moved to 2 December 2027 under the Digital Omnibus.
Why it matters for evidence: Logs must be kept for six months, and users must be told how AI shaped decisions that affect them.
Source: meilynx.com
- 03
Machine manufacturers face 2027 traceability rules
From 20 January 2027, the EU Machinery Regulation requires safety-relevant software changes to be logged and traceable for five years. Machines must also be protected against corruption.
Why it matters for evidence: Every software change to a safety-critical machine has to be traceable for five years, unchanged.
Source: axelent.com
- 04
Deepfakes in ads carry 3 per cent fine from August
From 2 August 2026, AI-generated faces, voices and deepfakes in advertisements must be disclosed, with a 3 per cent turnover fine for non-compliance.
Source: instagram.com
- 05
AI agent audit trails need the full context
AI audit trails must capture not just user inputs but the complete operational context: prompts, system instructions, data sources, tool calls, reasoning steps and outputs. That is what proves what the system actually did.
Why it matters for evidence: High-risk systems must log events automatically, and proving what an AI agent did means sealing the full context.
Source: quora.com
- 06
Enterprise AI agents need controls before production
Enterprise leaders must require comprehensive audit trails and emergency controls before allowing AI agents near production systems. Production readiness demands documented logging, human oversight and the ability to prove what autonomous agents actually did.
Source: linkedin.com