What RedFlag saw in Sam Bankman-Fried's 2022 DealBook interview
← All articles · By the RedFlag team · September 26, 2026
On November 30, 2022, three weeks after FTX collapsed, Sam Bankman-Fried joined the New York Times DealBook Summit by video link and told Andrew Ross Sorkin he “didn't knowingly commingle funds” and had not tried to commit fraud. We analyzed a nine-minute stretch where he is mostly on screen alone (30:00–38:50 of the NYT recording). RedFlag is not a lie detector — it measures where a speaker's voice and face diverge from their own baseline and leaves the interpretation to you.
The public record is settled. In November 2023 a federal jury convicted Bankman-Fried on seven counts of fraud and conspiracy, and in March 2024 he was sentenced to 25 years in prison; the conviction stands while he appeals. That lets us look honestly at what the measurable signals did in an interview about the collapse.
The analyzed video
RedFlag's analysis of the clip. The screen turns red during flagged moments, a green bar marks stretches within normal range, and the counter climbs as each flag is caught.
Overlay rendered from the analyzer's own output for this clip (same pipeline as the app's capture mode). Original clip: YouTube ↗.
What RedFlag flagged
Across the clip, RedFlag flagged 15 moments totalling 2 minutes 23 seconds. Each flagged episode lasts at least 3 seconds; shorter blips are dropped as measurement noise.
| # | Timestamp | Length |
|---|---|---|
| 1 | 0:00 – 0:06 | 6s |
| 2 | 0:23 – 0:29 | 6s |
| 3 | 2:05 – 2:17 | 12s |
| 4 | 2:49 – 2:54 | 6s |
| 5 | 3:11 – 3:17 | 6s |
| 6 | 3:26 – 3:31 | 5s |
| 7 | 3:48 – 3:52 | 4s |
| 8 | 4:57 – 5:02 | 5s |
| 9 | 5:28 – 5:48 | 20s |
| 10 | 6:26 – 6:33 | 7s |
| 11 | 6:45 – 6:56 | 11s |
| 12 | 7:11 – 7:16 | 5s |
| 13 | 7:26 – 7:46 | 20s |
| 14 | 8:09 – 8:14 | 6s |
| 15 | 8:27 – 8:52 | 26s |
What the signals did
| Signal | Flagged moments | Rest of the clip |
|---|---|---|
| Pitch variability (F0 semitone SD; normal band 2.3–4.5) | 3.18 st | 2.67 st |
| Blink rate | 11.11/min | 10.8/min |
| Voiced continuity | 0.28 | 0.3 |
This one is not a monotone story. Across the flagged moments Bankman-Fried's average pitch variability was actually higher than in the rest of the segment (3.18 vs 2.67 semitones): the flagged windows mix flat stretches below the 2.3–4.5 band with agitated bursts above it. RedFlag's scorer treats departures from the normal speaking band in either direction as deviation, which is why both show up. Blink rate and voiced continuity barely moved (11.1 vs 10.8 per minute; 0.28 vs 0.30). Part of the segment cuts to a split view with Sorkin; Bankman-Fried's video box remains the largest face, so he stays the scored speaker, but the shared audio track includes Sorkin's questions. Thresholds are public in the scoring method.
What this does and does not mean
It does not mean the analyzer detected a lie. Flattened pitch, blink changes and unusual voicing all have innocent causes — rehearsal, fatigue, a hostile room, a hard question. The honest reading is narrower: at these moments, this speaker's delivery deviated from their own baseline in a measurable, repeatable way. What that meant is a judgment, and here the public record has already made it.
Run your own. Paste any public YouTube interview with a single speaker into the analyzer — free, no signup, in your browser. RedFlag charts the signals; what they mean stays your call.
Open the analyzerSources
- Analysis run: RedFlag analyzer (capture pipeline), September 2026, on YouTube IyoGdwVIwWw. Figures are the analyzer's own per-slot output.
- Bankman-Fried's November 2, 2023 conviction and March 28, 2024 sentencing (summary ↗).
- Original footage: New York Times Events, DealBook Summit 2022.
- Scoring method and thresholds: How RedFlag scores a video.