AI Trustworthiness: Path to Mass Adoption in Critical Industries
summary
Claims that standards can make high‑risk AI certifiable by 2025 are backed with concrete work: ISO already has 30 published and 36 ongoing standards and aims for nearly 100 in coming years; the EU AI Act delegates technical detail to standards bodies; Numalis deploys formal methods at scale to verify robustness across input subspaces rather than just data points; a funded EASA project tests three aviation systems: ATC speech‑to‑text, collision avoidance, visual inspection and offers a SAMPLe tool for engineers to run free evaluations.
AI-written summary · about
chapters (15)
- 0:00AI trustworthiness introduction
- 1:12AI adoption challenges
- 4:01Brute-force AI safety limits
- 5:06Global AI regulations
- 6:31Standardization role in AI
- 9:03ISO/IEC AI compliance
- 13:24EU AI Act requirements
- 15:58AI robustness standards
- 18:56AI validation methods
- 21:51AI verification in aviation
- 23:37Iterative AI validation
- 24:44AI development mindset shift
- 26:30AI compliance tools and skills
- 27:34Formal AI validation methods
- 28:03Q&A
related talks
01h 07mNew Governance and Inclusiveness in AI Standardisation How Far Have We Gone?
29mISO's aim for standardization in machine learning
34mRisks of AI Risk Policy: Five Lessons
17mPreserving integrity and trust in AI
31mSecuring the Future: Navigating AI Vulnerabilities and Evolving Security Practices
33mAI Regulation in Europe - what's in the pipeline?
24mImplementing Robust AI Security for Enterprise Innovation
24mTechnical Tracks