Four stories, one theme: *whose evidence wins when something's questioned*
From ICO warnings to Swedish banks under DORA review, the week's news asks the same thing: can you show what actually happened?
- 01
ICO: autonomous AI doesn't shift responsibility
The UK Information Commissioner's Office says companies remain fully responsible for protecting personal data when using AI systems, however autonomous. Autonomy doesn't move accountability.
Why it matters for evidence: Under GDPR, you need to show how decisions were made. A log nobody can trust makes that harder.
Source: Compliance Week (subscription)
- 02
FI examines Swedish banks' AI attack defences
Finansinspektionen will review how Swedish banks, insurers and financial infrastructure firms prepare for AI-powered cyberattacks, under DORA's digital operational resilience requirements.
Why it matters for evidence: DORA requires logs protected against tampering, deletion and unauthorised access. Proving that protection is what FI will examine.
Source: breakit.se (in Swedish)Our explainer →
- 03
Air Canada liable for chatbot's wrong answer
A Canadian tribunal ruled Air Canada legally responsible for misleading information its AI chatbot gave a customer. The airline's claim that the chatbot was a separate legal entity was rejected.
Why it matters for evidence: The customer had a screenshot. The airline had no record to the contrary, so the tribunal assumed such evidence wouldn't have helped it.
Source: linkedin.comOur explainer →
- 04
No standard yet for AI agent audit trails
A critique argues no vendor standard yet meets regulatory audit requirements for AI agents, particularly SEC rules requiring reconstruction of original records. Organisations must own their audit trail strategy.
Source: ibl.ai
- 05
TÜV SÜD and QuantPi partner on AI conformity
TÜV SÜD and QuantPi will combine expertise to assess AI systems in regulated products under the EU AI Act. The partnership helps manufacturers prove their AI systems meet safety and compliance requirements.
Source: presseradar.de (in German)
- 06
Swedish watchdog finds influencer marketing still wanting
Konsumentverket reports continuing shortcomings in influencer marketing in 2026. Organisations hiring influencers must show advertising is properly marked and AI-generated content declared under the rules.
Source: collabri.se (in Swedish)