Deepfakes, fines and the courts: trust is becoming a technical question
OpenAI and Meta warn US judges about deepfakes. Bank of America pays $1.2 billion. Courts face AI-generated evidence they can't verify.
- 01
OpenAI and Meta warn US courts about deepfakes
OpenAI and Meta have told the US judiciary that deepfakes are now so convincing that determining whether audio and video evidence is real will be increasingly difficult.
Why it matters for evidence: Courts rely on free assessment of evidence. When a recording can be dismissed as potentially artificial, proving what actually happened becomes harder.
Source: reuters.com
- 02
Bank of America fined $1.21 billion for AML lapses
The US Treasury has fined Bank of America $1.21 billion for failures in anti-money laundering controls and alleged sanctions violations.
Why it matters for evidence: DORA requires financial firms to protect logs against tampering and unauthorised access. When controls fail, the question is what the monitoring systems actually did.
Source: Compliance Week (subscription)
- 03
Deepfake video of Bhutanese woman posted without consent
A pornographic deepfake showing a woman's face was reportedly posted online. No one can prove which AI system was used, when the manipulation happened, or who made it.
Why it matters for evidence: Article 50(4) requires deepfake creators to disclose that content is artificial. When nobody can prove who created it or when, the rule is hard to enforce.
Source: AI Incident Database
- 04
EU begins enforcing AI Act transparency rules
Article 50 of the AI Act is now in force, requiring AI systems to identify themselves and AI-generated content to be labelled. Applies across sectors where AI talks to people or creates public content.
Source: cineuropa.org
- 05
Academic paper asks who signs off on AI in audits
A new paper examines ISO 55001 asset management audits and the AI Act's logging requirements. It asks who bears legal responsibility for AI system behaviour before an auditor arrives.
Source: papers.ssrn.com
- 06
Real evidence dismissed as deepfakes in court
Deepfakes don't just create false evidence. They let real evidence be dismissed as AI-generated. Courts now see video, audio and documents challenged as potentially artificial.
Why it matters for evidence: When authentic evidence can be dismissed as AI-generated, proving what happened requires showing that a recording hasn't been tampered with since capture.
Source: instagram.com