Stage 3: independently tested.
The rule that placed itSomeone other than the maker has tested it: an independent benchmark, a peer-reviewed paper, or a hands-on test by a trusted outlet.
The evidenceOn deepfakes that actually circulated in 2024, open-source detectors' accuracy (AUC) fell by about half compared with older benchmarks; commercial detectors did better but not as well as human forensic analysts. Deepfake-Eval-2024, arXiv 2503.02857, May 27, 2026.
A stage says how far the evidence has got. It is not a rating of the AI and not advice.
- 1ClaimedNo evidence yet
- 2LaunchedNo evidence yet
- 3Independently testedPlaced here
- 4Tested with real peopleNo evidence yet
- 5Regulator-clearedNo approval process exists
- 6Widely used with a verified recordNo evidence yet
Everything we found.
In ladder order, newest first within each kind. Makers' own claims are labeled as theirs.
What independent tests found
- Deepfake-Eval-2024 authors (Chandra et al.)
On deepfakes that actually circulated in 2024, open-source detectors' accuracy (AUC) fell by about half compared with older benchmarks; commercial detectors did better but not as well as human forensic analysts.
“with AUC decreasing by 50% for video, 48% for audio, and 45% for image models compared to previous benchmarks.”
Deepfake-Eval-2024, arXiv 2503.02857 · May 27, 2026
Rules and approvals
- EU · Rule · European Union (AI Act)
The EU AI Act requires anyone using AI to make a deepfake image, audio or video to disclose that it was artificially generated or manipulated.
EU AI Act (Regulation (EU) 2024/1689), Article 50, AI Act Service Desk, European Commission · Jul 27, 2026
- China · Rule · Cyberspace Administration of China, with three other national bodies
Since 1 September 2025, China has required platforms to check files for hidden AI labels and to add a visible notice when content is identified as AI-generated.
Cyberspace Administration of China and others, Measures for Labeling AI-Generated and Synthetic Content · Mar 14, 2025
Not found in our sources yet: what the makers claim; what has launched; tested with real people; how many people use it. No regulator approves AI for this task, so the approval step doesn't apply.
Can AI do these yet?
Answer my health questions
Stage 4 of 6Tested with real people
Randomized trial: chatbot users made no better health decisions
Spot breast cancer on a mammogram
Stage 5 of 6Regulator-cleared
Transpara has been FDA-cleared since 2018, most recently in Aug 2026
Check a mole for skin cancer
Stage 5 of 6Regulator-cleared
EU Class III CE mark lets DERM clear benign cases alone (company)