CaraComp Insider

According to security researchers, a convincing deepfake now costs about a dollar thirty-three to produce. One photo and a sixty-second voice clip. That's all an attacker needs to fake someone's face Read the full article: https://www.caracomp.com/podcasts/platforms-rush-to-face-scans-to-fight-deepfakes-theyre-solving-the-wrong-problem

A ninety-five percent match score from a facial recognition system sounds like a slam dunk. But that number might be closer to meaningless — because it's not measuring what you think it's measuring. Read the full article: https://www.caracomp.com/podcasts/a-95-match-score-sounds-definitive-heres-why-it-might-mean-almost-nothing

Amazon's facial recognition system once matched twenty-eight members of Congress to criminal mugshots. That wasn't a hack. It wasn't a glitch. The system was running exactly as designed — ju Read the full article: https://www.caracomp.com/podcasts/a-95-match-score-sounds-like-proof-in-a-million-face-database-it-means-50-000

South Dakota's governor just signed a law making deepfake pornography a felony. Creating it, sharing it — both now carry serious prison time. And that single bill exposes a gap most people haven't not Read the full article: https://www.caracomp.com/podcasts/deepfakes-felony-law-in-south-dakota-raises-the-bar-for-photo-evidence

In 2023, roughly half a million deepfakes circulated online. By 2025, that number hit eight million. And according to U.N. Women, ninety-eight percent of them are non-consensual pornographic images ta Read the full article: https://www.caracomp.com/podcasts/deepfakes-hit-8-million-courts-still-cant-trust-the-evidence

N.A.T.O. ran a military exercise earlier this year where cybersecurity firm Reality Defender slipped deepfake media into a simulated battlefield scenario. Experienced military officials — people trained to assess threats Read the full article: https://www.caracomp.com/podcasts/a-95-confidence-score-falls-apart-if-the-media-was-faked-before-you-ran-the-matc

An A.I. chatbot looked at a real video of Benjamin Netanyahu sitting in a café — and declared it a hundred percent deepfake. It wasn't. And what happened next broke something fundamental about how we Read the full article: https://www.caracomp.com/podcasts/ai-called-netanyahus-caf-video-a-deepfake-it-wasnt-thats-the-real-problem

A verified identity profile just cleared every automated check. The credential is cryptographically authentic, government-issued, untampered. And there's still roughly a one-in-twenty chance the pers Read the full article: https://www.caracomp.com/podcasts/verified-doesnt-mean-matched-why-5-6-of-passed-identity-checks-still-hide-the-wr

Back in 2018, a research paper proved you could catch deepfakes because the A.I.-generated faces never blinked. That discovery spread everywhere. And it created one of the most dangerous blind spots Read the full article: https://www.caracomp.com/podcasts/deepfake-detections-biggest-mistake-one-tell-fools-investigators-every-time

An A.I. chatbot looked at a real video of Benjamin Netanyahu sitting in a café — verified footage, confirmed location, actual event — and declared it one hundred percent deepfake. Not fifty-fifty. Not suspicious. One hun Read the full article: https://www.caracomp.com/podcasts/netanyahus-caf-video-shows-why-i-saw-it-on-video-no-longer-counts-as-evidence

A detection tool says it's ninety-six percent accurate. In the lab, maybe. In the real world, that number collapses to somewhere around sixty-five percent. That's barely better t Read the full article: https://www.caracomp.com/podcasts/deepfake-detectors-promise-96-accuracy-in-the-real-world-they-drop-to-65

YouTube just handed politicians and journalists a fast-track button to flag and remove deepfake videos of themselves. But actors and athletes got that same button months earlier, back in December. So Read the full article: https://www.caracomp.com/podcasts/youtubes-deepfake-shield-for-politicians-changes-evidence-forever

Automated deepfake detection systems drop to about half accuracy when they're up against real-world fakes. And humans? We score barely better than a coin flip — around six in ten correct. That means Read the full article: https://www.caracomp.com/podcasts/deepfake-on-your-desk-how-smart-investigators-use-face-comparison-as-a-first-pas

YouTube just rolled out deepfake detection for government officials, political candidates, and journalists. Not as a research experiment. As a product feature. And it works a lot like Content I.D. — Read the full article: https://www.caracomp.com/podcasts/youtubes-deepfake-detection-tool-just-changed-the-rules-for-video-evidence

A deepfake can fool your eyes in a single frame. But it can't fake the way your jaw rotates across hundreds of frames. That's the difference modern detection tools actually exploit. Read the full article: https://www.caracomp.com/podcasts/how-deepfake-detection-actually-works-its-all-about-movement

A European court just ruled that the agency which pseudonymises a dataset — swaps names for codes, blurs identifiers — still bears full G.D.P.R. obligations for that data. Not reduced obligations. Fu Read the full article: https://www.caracomp.com/podcasts/blurring-a-name-doesnt-anonymise-a-face-what-gdpr-actually-says

A European grocery chain got hit with a major fine for running facial recognition on shoppers without meeting biometric data rules. That penalty didn't happen under some future law. It happened under Read the full article: https://www.caracomp.com/podcasts/eu-digital-omnibus-will-redraw-the-rules-on-biometric-evidence

A grandmother in Tennessee spent six months in jail. The facial comparison algorithm didn't send her there. An investigator who treated a machine's suggestion like a verdict did. Read the full article: https://www.caracomp.com/podcasts/ai-face-match-probable-cause-a-grandmother-paid-the-price

A deepfake costs about ten bucks to make. Defeating three independent biometric sensors at the same time? That pushes into nation-state territory. The gap between those two numbers explains why multim Read the full article: https://www.caracomp.com/podcasts/multimodal-biometrics-why-face-fingerprint-voice-defeats-deepfakes

What if the face that breaks your investigation isn't from a criminal — but from a machine? A synthetic face built by A.I. can now mimic the exact skin texture, landmark spacing, and micro-asymmetry that trained examiner Read the full article: https://www.caracomp.com/podcasts/how-to-stress-test-your-facial-comparison-method-against-deepfakes

A grandmother in Tennessee was sent to jail. Facial recognition software said she was a match. She wasn't. And the investigators who arrested her never bothered to check. Read the full article: https://www.caracomp.com/podcasts/ai-facial-recognition-sent-an-innocent-grandmother-to-jail

A facial recognition algorithm just scored near-perfect accuracy on a national benchmark. Then an investigator ran it on real surveillance footage. And the results weren't even Read the full article: https://www.caracomp.com/podcasts/nist-benchmarks-are-impressive-heres-what-they-dont-tell-investigators

Here's something wild. A facial recognition system rated ninety-nine percent accurate can still miss one out of every hundred genuine suspects. And that's under perfect lab conditions. Read the full article: https://www.caracomp.com/podcasts/what-99-accurate-facial-recognition-actually-means-for-your-case

Here's something that trips up even experienced investigators. The facial recognition failures you see in the news — they're real. But they're describing a completely different problem than what most Read the full article: https://www.caracomp.com/podcasts/the-face-recognition-error-thats-wrecking-investigations

Imagine a defense attorney turns to you in court and says, "Walk me through exactly how you compared those two faces." Step by step. With documentation. Would you feel confident? Read the full article: https://www.caracomp.com/podcasts/law-enforcement-isnt-abandoning-face-tech-its-regulating-it