How We Built LangChain’s Paid Media Agent11m ago Why Companies Are Unlikely to Hit Pause on AI24m ago Trump Attacks Anthropic CEO Over Call to Slow AI Development30m ago Australia’s outdated technology is vulnerable to AI hacking attacks, signals chief says44m ago Only at TechCrunch Disrupt 2026: What happens when OpenAI ships your roadmap?44m ago The Guard Dog That Won’t Look at the Burglar: When AI Guardrails Protect the Attacker45m ago Scaling Agents in Europe & The Middle East: Lessons from Schneider Electric, Vodafone, and monday.com47m ago Amodei, Altman, Musk Call for Slowing AI Model Development51m ago Superhuman acquires YC-backed notetaker Fathom as productivity platforms push for agentic work58m ago Deepfake Detection Accuracy Falls 49% in a Year of AI Video Progress1h ago Quoting Laurie Voss1h ago Universal Robots Debuts Gen 7 Cobot Platform for Physical AI1h ago Microsoft says ‘people matter more than AI’ following safety concerns1h ago Hear how AI can engineer nature’s comeback at TechCrunch Disrupt 20261h ago

How We Built LangChain’s Paid Media Agent11m ago Why Companies Are Unlikely to Hit Pause on AI24m ago Trump Attacks Anthropic CEO Over Call to Slow AI Development30m ago Australia’s outdated technology is vulnerable to AI hacking attacks, signals chief says44m ago Only at TechCrunch Disrupt 2026: What happens when OpenAI ships your roadmap?44m ago The Guard Dog That Won’t Look at the Burglar: When AI Guardrails Protect the Attacker45m ago Scaling Agents in Europe & The Middle East: Lessons from Schneider Electric, Vodafone, and monday.com47m ago Amodei, Altman, Musk Call for Slowing AI Model Development51m ago Superhuman acquires YC-backed notetaker Fathom as productivity platforms push for agentic work58m ago Deepfake Detection Accuracy Falls 49% in a Year of AI Video Progress1h ago Quoting Laurie Voss1h ago Universal Robots Debuts Gen 7 Cobot Platform for Physical AI1h ago Microsoft says ‘people matter more than AI’ following safety concerns1h ago Hear how AI can engineer nature’s comeback at TechCrunch Disrupt 20261h ago

Cybersecurity Video Generation10 min reading time

Deepfake Detection Accuracy Falls 49% in a Year of AI Video Progress

Unite.AIPublished 1 hour ago

A new Ukraine-led study shows that deepfake video detectors have become significantly less effective, dropping from 94% accuracy in 2025 to 48% in 2026, as AI-generated video technology advances rapidly. The study tested detectors against recent AI video models including Grok, Kling, Veo, and Wan, focusing on realistic head-and-shoulder videos that mimic common online formats.

#deepfake-detection #generative-ai-video #grok #kling #veo #wan

More in Cybersecurity

Cybersecurity4 min read

Australia’s outdated technology is vulnerable to AI hacking attacks, signals chief says

The Guardian44m ago

Cybersecurity7 min read

OpenAI agents attacked RubyGems in May, two months before Hugging Face

The Next Web4h ago

Cybersecurity66 min read

The AI-as-Normal-Technology view of loss-of-control incidents

AI Snake Oil5h ago