Courts, cameras and the evidence problem with AI
Four stories, one theme: showing what happened matters more when AI is involved.
- 01
Lawyer fined for ChatGPT hallucinations
A US lawyer and colleagues were fined $5,000 for submitting fake case citations generated by ChatGPT. They couldn't verify what the AI had produced before relying on it.
Why it matters for evidence: Free evidence rules mean anything can be used, but weight depends on how trustworthy it is.
Source: spellbook.com
- 02
Deepfake video earns terminating sanctions
A California court found two video exhibits were deepfakes and imposed the harshest sanction on the party that submitted them. Full metadata and provenance now matter in court.
Why it matters for evidence: Courts scrutinise digital evidence more closely. Metadata and an unchanged record help prove what your systems recorded.
Source: truescreen.io
- 03
Product liability directive transposition due
EU member states must transpose the new Product Liability Directive within three months. Software and AI become products with strict liability, raising the bar for evidence.
Why it matters for evidence: After 9 December 2026, you may need to show what your software did on a given day to defend against liability claims.
Source: linkedin.com
- 04
Predictive hiring software faces high-risk rules
A recruitment assessment tool using AI falls under the AI Act's high-risk requirements for employment. That means logging, human oversight and a right to explanation from 2 December 2027.
Source: assessio.com (in Swedish)