OpenAI tightens Sora deepfake safeguards

Rizzle

Watch OpenAI tightens Sora deepfake safeguards - Rizzle on Dailymotion

Transcript

00:00OpenAI Tightens Sora Deepfake Safeguards

00:04OpenAI's newest toy, Sora 2, is like Photoshop on steroids, if Photoshop had a moral hangover.

00:10The AI video generator has proven so good at faking reality that it's already spitting

00:15out shockingly convincing deepfakes of everyone from Martin Luther King Jr. to SpongeBob.

00:21And while it might sound like an edgy art experiment, users are discovering that Sora

00:25can also turn them into unwilling actors in racist rants or fetish videos.

00:30Inside the app, users at least know everything's fake, but once those clips start circulating

00:34on TikTok or Instagram, all bets are off.

00:37OpenAI's watermarking and authenticity tools might as well be invisible ink.

00:42The company proudly points out that every Sora video includes C2PA metadata, a sort of digital

00:47nutrition label meant to prove how a video was made.

00:50Unfortunately, almost no one can see it.

00:53C2PA, or content credentials, was supposed to be the internet's secret weapon against

00:58deepfakes.

00:59Adobe, OpenAI, Google, Meta, and even the US government have all backed it.

01:05In theory, it's a great idea, invisible metadata showing who made what, when, and how.

01:09In practice, nobody checks.

01:12Labels are either missing, mislabeled, or hidden behind collapsed menus.

01:17That's awkward for OpenAI, which sits on C2PA's steering committee while simultaneously

01:22pumping out a tool that makes fake videos of real people shouting white supremacist slogans.

01:26Deepfake watchdogs say Sora 2's identity filters were cracked within 24 hours of launch.

01:32Adobe's team insists things are improving, slowly.

01:35People need clear information about how content is made, said Andy Parsons, Adobe's content

01:41authenticity lead, sounding like someone watching the Titanic sink while complimenting the band.

01:45Experts agree, C2PA alone won't save us.

01:50Metadata can be deleted, watermarks removed, and no one's really enforcing the rules.

01:54Lawmakers are now pushing anti-deepfake bills to catch up.

01:58Until then, AI creators like OpenAI are both arsonists and firefighters building the deepfake

02:03engines while selling us the smoke alarms.

02:155 girls are coming up.

02:19Be brave.

02:21Be brave.

02:305,060.

02:3210,258.

02:337,808.

02:3410,369.

02:3411,

02:34eight, eight, eight, eight.

02:3510,7.

02:3611,2,

02:37seven.

02:3811,9.

02:3910,9.

02:4010,5,

02:4211,9.

02:4311,9.

Category

Transcript

Be the first to comment

Recommended