Towards a Safe Harbour for Public Interest AI Research
summary
Panel argues that independent audits require concrete fixes: structured, multi‑tiered researcher access; legal safe harbours beyond voluntary clauses; and tooling for safe sharing. Evidence cited includes a 4,000% increase in health data breaches, stale research feeds like a 3-day delay in research API data, and industry pilots such as Anthropic’s Cleo aggregated usage tool. Current code of practice and vendor gatekeeping leave researchers blocked, APIs throttled or revoked, and bias, hallucination and provenance questions unresolved.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 5:25Health AI research obstacles
- 15:59Need for researcher access
- 22:35AI models vs social media
- 24:23Researcher access programs
- 33:55Corporate transparency arguments
- 40:02Legal merits of transparency
- 43:15Technical hurdles in AI
- 48:10Defining safe harbor
- 51:45Policy and company actions
- 1:01:24Audience questions and responses
related talks
01h 16mLifting the Hood on Big AI - The Future of Transparency and Accountability in AI
01h 18mThe Role of Research and Researchers in AI Governance
53mWild West's End
28mWhy Private AI Matters: Data Sovereignty as the Foundation for Trustworthy AI on the Open Web
18mDeveloping AI in a privacy conscious world
44mAI, Corporate Responsibility & Democratic Legitimacy
01h 08mTHE GOVERNANCE OF AI: CONVERGENCE OR DIVERGENCE?
01h 09mGOVERNANCE AND REGULATION OF AI FROM THE PERSPECTIVE OF AUTONOMY AND PRIVACY