Responsibility in Depth: Layering Licensing, Regulation, and More
summary
Argues that copyright licenses alone fail for open ML, using the RAIL license adoption by Stability — recent press shows models used for non‑consensual porn — to show four concrete failures: vague definitions, unknown downstream users, no enforcement budget, and cross‑jurisdiction limits. Proposes layering defenses: Swiss Cheese model of licensing plus community codes of conduct and forum enforcement, employee power and strikes, systematic red teaming and QA, data trusts and contractual agreements for verifiable use, public‑benefit orgs, and targeted litigation to make rules stick.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:51Open source in 1997
- 7:23Licensing fails for AI
- 12:58Swiss Cheese safety model
- 17:56Layering governance tools
- 23:15Data governance for public benefit
- 26:29Enforcement strategies
- 28:55Legislation and community input
- 30:40Imagination in governance
related talks
29mDevOps Days MTL 2019 - Richard Fontana
26mPanel Discussion: Open Source Licensing
22mEthical Algorithms, Licensing Impasses: The Intersection of Free Software and AI openness
25mWhat Makes Open Source Work: Licensing and Beyond
01h 38mOpen Source Licensing: Types, Strategies and Compliance
43mSoftware Licensing and Compliance: It's All About Community
48mWhat You Need to Know About Open Source - Trust Me, I'm a Lawyer
15mEnough Foundations Already!