Decoder with Nilay Patel

That UL safety logo is a lot more complicated than it looks

1h 2min

UL Solutions CEO Jennifer Scanlon discusses the complex and evolving role of the UL safety logo. Originally focused on electrical safety for insurance, UL now tackles new challenges like lithium-ion battery fires in e-bikes and the non-deterministic nature of AI systems, for which they've introduced the UL3115 standard. Scanlon highlights UL's transition to a for-profit entity, its global operations, and the ongoing efforts to maintain trust and relevance in a market flooded with uncertified products, while navigating political challenges like the cybertrust program controversy.

Summarized by Podsumo

🎙️ Ask about this episode

✨ Key Takeaways

1

Evolving Safety Landscape: UL has expanded its scope from traditional electrical safety to critical new areas like lithium-ion battery safety (e.g., e-bikes) and AI systems, introducing the UL3115 standard for AI evaluation.
2

Tangible Impact of Standards: The adoption of UL's e-bike battery safety standards in New York City resulted in a 75% reduction in related fires, showcasing the direct benefit of certification.
3

Hybrid Business Model: UL Solutions operates as a for-profit company but is intrinsically linked to its non-profit UL Standards & Engagement and UL Research Institute, ensuring its commitment to public safety science.
4

Marketplace Safety Challenges: The proliferation of uncertified products on platforms like Amazon, particularly cheap lithium-ion devices from China, poses significant safety risks and complicates enforcement.
5

AI Testing Complexity: Evaluating AI systems is challenging due to their non-deterministic nature; UL3115 addresses this by focusing on 200 criteria related to data transparency, algorithmic bias, and responsible development.

💬 Notable Quotes

"Innovation without safety is failure."
"We have been very transparent about our operations, our relationships, all over the world, and we continue to be so."
"The more that it goes into the training of the models, the more difficult that you have in trusting if those models have the capacity for the future."