Deepfakes Threaten Entertainment Industry with $40B Losses

Tom Cruise became an early viral example of how convincing deepfakes can be. Today, they’re in the top 3 most mainstream threats we see on social media. Anyone with a motive and minimal skill can generate believable impersonations of actors, executives, or entire marketing campaigns. Fake endorsements, hoax videos, explicit content and synthetic “statements” are stupid easy to create. Here’s a statistic that should shock you: Losses to AI-related fraud are forecast to reach $40B in the US alone. Hollywood has always adapted to new technology, but social media threats are evolving faster than any previous disruption, and AI is accelerating both the scale and sophistication of attacks. If you work with entertainment brands or celebrities, our report on the biggest threats on social media in the entertainment industry is a must-read: https://lnkd.in/dsdXAdMK

The deepfake threat didn't suddenly get worse this week — but a single viral demo just forced a major platform to slam the brakes on its new AI video tool. Here's why that matters for every investigator reading this. A content creator used ByteDance's new generative video model to recreate his own voice and body from a single photo. He called the results "terrifying." Within 72 hours, the platform restricted the tool entirely. That speed tells you everything. We've crossed a threshold where a single demo can collapse the distance between "cool feature" and "evidence crisis" in days, not years. Now here's what most people are missing. The real problem isn't the fakes themselves. It's what researchers call the "liar's dividend" — the moment where even authentic evidence becomes suspect simply because convincing fakes exist. A California Superior Court judge recently identified a plaintiff-submitted witness video as AI-generated after noticing unnatural facial movements and inconsistent metadata. That's a judge catching a fake in real time. But flip it around. What happens when legitimate video gets thrown out because opposing counsel argues it could be fabricated? Federal courts are already wrestling with this. The U.S. Judicial Conference considered deepfake-specific amendments to evidence rules and kept them in reserve — not rejected, but not adopted either. Proposed changes include specialized authentication processes and new standards for machine-generated evidence. Until those settle, investigators face inconsistent rules across every jurisdiction they work in. The investigator's playbook has to shift from "detect the fake" to "prove the real." That means chain-of-custody documentation, metadata preservation, and facial comparison analysis that produces court-ready reporting — not screenshots from consumer tools that wouldn't survive a motion to exclude. This is exactly why we built CaraComp the way we did. When the burden shifts to proving evidence is authentic, the quality of your analysis tools stops being a convenience and becomes the difference between admissible and dismissed. When a client hands you a "smoking gun" video in 2026, what will your default assumption be — real until disproven, or fake until validated? I dropped the full article in the comments. #DeepfakeEvidence #DigitalForensics #PrivateInvestigator #OSINTCommunity #InvestigationTech

Keeping people safe online is an ever-evolving challenge. As bad actors use AI to create more complex and deceptive ads, we are fighting back with our own advanced AI to stay steps ahead. Today we released our 2025 Ads Safety Report – and the impact of our Gemini-powered tools is clear: 🚫 We blocked or removed 8.3 billion malicious ads in 2025 ✋99% of those were stopped before anyone could see them And because our AI handles this scale with speed and precision, our safety experts are freed up to focus on the most complex cases that require real human judgement. Learn more about how we are tackling adversarial threats and protecting our ecosystem.  Link in comments 👇

In the early hours of Friday morning, Daniel Moreno-Gama was arrested in connection with an attack on Sam Altman's San Francisco home and the nearby OpenAI headquarters. We took a deep dive into the online presence of Moreno-Gama, combing his social media accounts and Substack. What we found paints a picture of a young man increasingly focused on AI and the "existential threat" it poses. Read more about how he went from having a fear of dying to an AI "martyr" in Business Insider https://lnkd.in/eGEeCskh

YOUR PROFILE AI systems can profile you in 8 minutes of TikTok. Map your vulnerabilities. Predict your political leanings. Target you with deepfakes designed to trigger fear or anger. TikTok's own research: "We can predict susceptibility to emotional manipulation with 68-74% accuracy." Meta's advertiser docs: "Emotional resonance targeting" = 8.7x higher conversion rates. The scale: 500,000 deepfakes in 2023 → projected 8 million by 2025. The defense: STRIDE framework — detection, regulation, inoculation, coordination. Reduced disinformation spread by 70% in simulations. Business reality: If platforms can micro target your customers with emotionally manipulative content, who's protecting your brand from the same attacks? Cognitive warfare is here. Prepare or react. #AI #STRIDE #warfare

YouTube Bans Pro-Iran AI “Lego” Propaganda Channel Mocking President Trump, Exposing Digital Warfare Against America. YouTube has removed a pro-Iran channel producing viral AI-generated Lego-style videos that mocked President Donald Trump highlighting the growing threat of foreign propaganda targeting the United States. The channel which gained millions of views, used humor and artificial intelligence to spread anti-American narratives to younger audiences. READ FULL NEWS STORY: https://lnkd.in/ddY36uvn

YouTube Bans Pro-Iran AI “Lego” Propaganda Channel Mocking President Trump, Exposing Digital Warfare Against America. YouTube has removed a pro-Iran channel producing viral AI-generated Lego-style videos that mocked President Donald Trump highlighting the growing threat of foreign propaganda targeting the United States. The channel which gained millions of views, used humor and artificial intelligence to spread anti-American narratives to younger audiences. READ FULL NEWS STORY: https://lnkd.in/ddY36uvn

“Around 63% of iGaming fraud comes from bonus abuse, and roughly 10 to 20% of marketing spend is lost to it.” 🗣️ In a fireside chat at the iGB@ICE Studio, EveryMatrix CEO Marc Burroughes, speaks on fraud and bonus abuse. EveryMatrix specialises on the topic and are utilising AI and targeted approaches for prevention. Listen to the full discussion here for more insights: 👇 https://lnkd.in/ejmnFUYQ

Criminals are becoming increasingly sophisticated and creative in their efforts to steal money and deceive through technology, targeting people on social media and using artificial intelligence (AI) to create highly convincing scams and impersonations. Recent research found that someone lost money to a fraudulent seller on Facebook or Instagram every six minutes in the UK in the first half of last year, with many saying it’s the fastest-growing type of fraud in our region. There’s no question that it remains a top focus area for all financial institutions. Please read the full article here: https://lnkd.in/eteJXQ3A

An Australian university ran a four-week war game where 270 students built AI bots to manipulate a fake social media platform and swing a simulated election. The bots succeeded — shifting the vote 1.8 percentage points, enough to change the outcome. The experiment was inspired by real attempts to poison AI chatbots with propaganda ahead of Australia's 2025 federal election and Chinese bot campaigns targeting Australian referendums. The lesson: AI-generated influence content is cheap to produce at industrial scale, and current social media platforms are not equipped to detect or remove it fast enough. 🎓 #CyberNewsLive https://lnkd.in/gQbn5kKA