GPAI Code of Practice & Systemic Risk
summary
Panel explains that the GPAI Code of Practice turns EU GPAI rules into concrete steps: mandatory safety framework since 2023 and public model reports (system cards) for tested risks, a one‑year phase‑in before sanctions, and startup relief narrowed to 10 commitments for startups. Key disputes are the compute threshold debate that decides which models are “frontier,” the AI Office’s triple regulator/standard‑setter/enforcer role (target ~40 staff vs UK’s ~200), and downstream firms’ poor awareness of how to use the code.
AI-written summary · about
chapters (12)
- 0:00GPAI Code of Practice intro
- 1:26Safety and security overview
- 2:30Industry perspective on the code from OpenAI
- 5:48FLI critique of the code
- 9:23Germany’s government view
- 13:46Impact on AI value chain
- 17:37Implementation challenges
- 23:35AI Office’s triple role
- 28:11AI Office oversight capabilities
- 31:44Brussels effect on AI governance
- 36:24Global governance and competitiveness
- 38:50Closing remarks
related talks
01h 20mFLOPs and beyond Decoding the AI Act’s Systemic Risk Criteria
01h 10mNavigating the Interface Between the AI Act and GDPR Combining Innovation and ..
01h 15mWhen Is a Summary of Data Used to Train GPAI Sufficiently Detailed?
28mAI Act - What is important to know from a developers perspective?
01h 17mBias, Discrimination and Data Protection in the Age of GPAI
36mResponsible AI: In Theory, Practice, and Policies
01h 11mTransforming GDPR into a Risk Based Harm Tool Alongside Specific AI Regulation...
01h 08mTHE GOVERNANCE OF AI: CONVERGENCE OR DIVERGENCE?