The Bunker – News without the nonsense Law & Order: Artificial Intelligence Unit – How do we police A.I.?
Mar 12, 2026
Dr Federica Fedorczyk, Research Fellow and AI ethics specialist studying AI’s impact on criminal justice. She explores a wide range of AI harms from deepfakes and election interference to surveillance and cyberattacks. Discussion covers mapping foreseeable misuse, platform and product liability, regulatory choices across jurisdictions, and how non-regulatory tools like courts and social pressure shape outcomes.
AI Snips
Chapters
Transcript
Episode notes
AI Enables Many Distinct Harms Beyond Deepfakes
- AI enables a wide range of illegal harms beyond deepfakes, from election interference to automated cyberattacks.
- Federica Fedorczyk lists examples: mass surveillance, militarised automation, discrimination by emotion, and behaviour manipulation of minors.
Non‑Consensual Sexual Deepfakes Are Mass Scale
- Non-consensual sexual deepfakes are widespread and primarily victimize women and children.
- Federica notes nearly 3 million sexual deepfakes were created over 10 days on a major public platform, showing this is mainstream behaviour not fringe.
Anticipate Misuse And Build Guardrails From Design
- Map and anticipate both short‑term and long‑term misuse when designing AI systems.
- Federica urges researchers to predict risks (deepfakes existed since 2017) and build guardrails from the design stage onward.