What happened
ETSI published TS 104 275, developed by its Technical Committee on Securing Artificial Intelligence (TC SAI), organizing safeguards around five lifecycle stages: secure design, development, deployment, maintenance and end of life.
Why it matters
The specification sets out content filtering, labeling, monitoring and human oversight, plus test scenarios and pass/fail criteria, helping AI organizations address deepfake risks at the design stage rather than detecting manipulated content after it is created.
WHO IT HITSAI developers and operators building or running generative AI systems are the direct audience, since the specification gives them requirements and pass/fail test criteria for demonstrating that deepfake safeguards are in place.
Summaries like this, every morning.
The specification reflects a shift in how deepfake risk is framed. Rather than treating detection of manipulated content as the main defense, ETSI TS 104 275 repositions governance and deepfake mitigation as a continuous process to be addressed throughout the AI system's lifecycle. That is why its requirements are organized around secure design, development, deployment, maintenance and end of life, and why it includes risk assessments, documentation and human oversight rather than only after-the-fact detection.
The document also connects to work already in place. It builds on baseline cybersecurity requirements set out in ETSI's EN 304 223 standard for AI models and systems, and it supports objectives laid out in the EU AI Act. In the release, Alex Cadzow, rapporteur and secretary of ETSI TC SAI, noted that harmful, misleading and illegal deepfake material can be created from the misuse of legal and commercially available software, and said AI organizations need support in addressing the risks much earlier, starting with how generative AI systems are designed.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
