Resemble AI ads

Deutsche Telekom just made the move every service provider should be making. They're demonstrating how Resemble AI’s real-time deepfake detection secures AI Agents at the network level. We're proud to ensure the secure infrastructure layer everyone else is missing. AI agents are about to handle your banking calls, your insurance claims, your healthcare appointments, so they'll talk to customers and to each other. And right now, most of that communication has zero verification, meaning there is no real way to tell whether the agent is legitimate or detect if a synthetic voice is impersonating a real person (or another AI agent). What DT understood before most: when you deploy AI agents at scale, you need to protect every conversation type. Agent-to-agent Agent-to-human Human-to-agent Human-to-human Every voice interaction needs to be verified in real time, across languages, across providers, compliant with EU AI Act requirements before enforcement even begins. That's what we're building together, detection that works at the network level, not security bolted on after the fact. If your service provider handles sensitive customer data and isn't thinking about AI agent security yet, it's time to ask them why. The companies building this way are futureproofing and everyone else is already behind. We're at MWC Barcelona this week in the Deutsche Telekom booth in Hall 3 -- come find us if you want to see what this looks (and sounds) like in real life. Massive thanks to the Deutsche Telekom team for the opportunity and the alignment to move fast on this mission: Abdu Mudesir , Ahmed Hafez , Alvaro Infantes , Ulrich Klenke , Dominik Schnieders and the many others!

Three stories are circulating this week that show how far we are into the global deepfake crisis. The UK is amending its Crime and Policing Bill to force platforms to remove non-consensual intimate images (including AI-generated deepfakes) within 48 hours or face being blocked entirely. PM Starmer called it a "national emergency." In New Delhi, PM Modi told the India AI Impact Summit that deepfakes and fabricated content are "destabilizing open societies" and called for global watermarking standards and authenticity labels on all AI-generated content. And in Germany, ZDF's flagship news program heute journal accidentally aired a Sora-generated video of an ICE arrest that never happened, in the same segment where the anchor warned viewers about fake videos online. 3.6 million people watched it - ouch. The are the same problem, at scale, causing damage in different ways. Non-consensual deepfakes destroy individual lives. Synthetic media slipping into newsrooms erodes collective trust in what's real. And heads of state are now standing at podiums saying what technologists have been saying for years: we need authentication infrastructure built into the content layer itself. The EU AI Act. India's IT Rules Amendment. The UK's Online Safety Act expansion. This is no longer a regional conversation. Detection, watermarking, and provenance aren't nice-to-haves, they're becoming legal requirements, globally, and fast. Real-time deepfake detection will be standard within 12 months, the only question is how much damage we accept before then.

$1 billion in suspected fraudulent home loans at Commonwealth Bank using AI-generated documents. Let that sink in. If you're not thinking about deepfake detection, you're already behind.

We officially launched Resemble Agent Detection today, so you can actually tell if AI agents are on your site or in your product. Most bot detection leans on network signals like IP reputation, VPNs and browser fingerprints, and agents can fake all of those. Agent Detection runs on System One, a model that reads real interaction behavior, like how a visitor moves, clicks and operates, so it detects agents even when they’re using a browser or disguising their origin. We tested Cloudflare, F5, Akamai and others against agents disguised as human visitors, and none of them caught more than 15%. Agent Detection caught 98.8% in the same test. The traffic light CAPTCHA had a good run, but no one will be sorry to see it go. Your first 1M Agent Detection sessions are free to try, if you want notifications or routing just let us know: https://app.resemble.ai/

One generative AI tool was responsible for 190 sexualized images per minute for a full eleven day run.

Fraudsters stopped asking for money transfers and started trying to get hired. Now the goal of bad actors using faceswaps and fake IDs is to be on your payroll.