What You Need to Know (And Why You Should Care) About AI Governance - Michelle Frost - NDC Oslo 2026

Value of the Information & Strength of the Argument

The talk provides a clear, actionable framework for integrating AI governance into product development. The speaker successfully bridges abstract ethical principles and legal requirements with concrete engineering decisions, using vivid examples and analogies. She strengthens her argument by referencing authoritative sources like NIST and the EU AI Act, and by acknowledging the dynamic nature of regulations. The emphasis on the shift in governance unit from the model to the entire workflow is particularly insightful, addressing a gap in many current practices. However, the argument would benefit from deeper technical detail on how to implement the proposed measures, and the rapid updates in legal information might challenge the talk’s longevity.

Scientific Rigor, Source Quality, Title Accuracy

The talk exhibits strong scientific rigor by citing established frameworks and regulations, including NIST AI RMF, GDPR, and the EU AI Act. The speaker distinguishes between legal advice and risk literacy, appropriately positioning her content as an engineering perspective. She explicitly references Olivia Gamblin’s book ‘Responsible AI’ and the International AI Safety Report 2025, lending credibility. The title is an accurate descriptor of the content, matching the talk’s scope and purpose. The presentation is well-structured, though the legal section is intentionally shallow due to variability, which is acknowledged. Overall, the sources are reliable and appropriately used.

The title perfectly matches the content: the speaker explains what developers need to know about AI governance and why it matters, covering ethical, legal, and technical dimensions.

8/10

High credibility due to systematic references to established frameworks (NIST, EU AI Act, GDPR) and the speaker's practical experience, though the speaker explicitly notes she is not a lawyer, which limits legal precision.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction, audience poll on team roles, and speaker's background in ML and ethics.
• 05:00Personal journey into fairness and AI, leading to the idea that ethics must be coupled with governance.
• 10:00Overview of the AI ethics landscape with six pillars and their subcategories.
• 20:00Discussion of the value alignment problem and the WhatsApp vs Signal example.
• 30:00Legal landscape: GDPR, EU AI Act timeline changes, and US state regulations via IAPP.
• 40:00NIST AI Risk Management Framework explained: govern, map, measure, manage.
• 50:00Practical framework synthesis and conclusion emphasizing engineering responsibility.

• NDC Conferences — Conference organizer and publisher of the talk.
• NDC Oslo — Event where the talk was recorded.
• NIST AI Risk Management Framework — Referenced as the primary framework for AI risk management.
• EU Artificial Intelligence Act — Discussed for its cumulative obligations and timeline changes.
• General Data Protection Regulation (GDPR) — Mentioned as a global data protection standard.
• IAPP State Regulation Tracker — Recommended for tracking US state AI regulations.

• NIST AI Risk Management Framework — Aligned with the talk's recommendation to use NIST for practical governance.
• EU Artificial Intelligence Act — Matches the talk's explanation of regulatory obligations and timelines.
• General Data Protection Regulation (GDPR) — Referenced as a baseline for data protection, consistent with the talk.