Courts ask what AI actually did, and logs become evidence
Five cases and one federal review this week, all asking the same question: can you show what happened, or only what you remember?
- 01
Airline liable for chatbot's invented refund rule
A Canadian tribunal ruled Air Canada responsible for incorrect refund information its chatbot provided. The airline argued the chatbot was a separate entity; the tribunal disagreed.
Why it matters for evidence: The customer had a screenshot. The airline had no record of what the chatbot actually said.
Source: instagram.com
- 02
US federal review asks who proves what AI did
A federal advisory committee is examining AI evidence on two tracks: reliability of AI conclusions and authenticity of AI-produced material. The review raises questions about who must prove what a system did when its output becomes evidence.
Why it matters for evidence: Courts will want to see what the AI actually did, not what the operator says it probably did.
Source: papers.ssrn.com
- 03
How should judges handle deepfake defences?
An article examines procedural challenges when an accused claims video evidence is AI-manipulated. It explores how courts assess such assertions under free evaluation of evidence.
Why it matters for evidence: When someone claims manipulation, the side that kept the original has the stronger case.
Source: linkedin.com
- 04
AI-generated content complicates divorce cases
Courts require parties to prove authenticity before AI-generated content can be considered as evidence in divorce proceedings. AI manipulation makes it harder to verify whether digital evidence reflects what actually occurred.
Source: thakkarfamilylaw.com
- 05
AI-generated audio bomb threat prompted evacuations
Officials assessed a Romanian-language audio bomb threat as AI-generated, which led to employee evacuations and security searches. The incident shows risks when synthetic audio goes undetected or unlabelled.
Source: AI Incident Database
- 06
Agent tampering: when monitors are silenced
Agent tampering is unauthorised interference with monitoring agents to hide activity or evade detection. Protecting audit logs from manipulation matters for regulations requiring tamper-proof logging.
Source: nhimg.org