‘Leveraging AI for Transforming Product Safety Learnings’
Plenary session 4: ‘Leveraging AI for Transforming Product Safety Learnings’ featured Elliot Kaye, a partner at Cooley and the former chairman of the US Consumer Product Safety Commission (CPSC), along with speakers from Amazon and Google.
The panel explored how companies can use artificial intelligence (AI) to transform product safety evaluations and data into actionable insights. Key points discussed included:
1. Current uses of AI in the product safety world
The panel identified that companies hold a significant amount of product safety data, but these data remain largely untapped due to their volume and unstructured nature. The processing of these data by AI systems (e.g., for trend analysis) – though not without risks – was identified as a key current use which allowed companies to design safer products, and evaluate and respond to product safety incidents more quickly and efficiently.
2. Moving from reactive to proactive to predictive applications
The panel considered the current pre-market and post-market surveillance applications of AI, before exploring its potential uses in the next three to five years. This included a review of the use of AI models in product design, material selection and manufacturing, before considering post-market surveillance activities, including the use of AI for multimodal analysis of consumer feedback. They outlined how companies could streamline processes by using AI models to select appropriate materials within the product’s design constraints, catch potential defects during the manufacturing process and comprehensively review consumer feedback.
Looking forward, the panel emphasised the potential for predictive use of AI to assist with decision-making processes, allowing for personalized safety for individual devices, including self-correcting mechanisms.
3. Not just for companies
The panel also discussed the use of AI by regulators around the world, and, in particular, the CPSC. The CPSC is currently using AI systems to scrub large data pools, such as consumer complaints, with an increased level of sophistication. The panel emphasised the importance for companies to keep up with regulators to avoid being on the back foot – for example, if a regulator’s use of AI led to better information about incidents involving the company’s products.
4. Potential issues
The audience identified that the most significant potential concern with using AI tools currently was the reliability and accuracy of the provided insights.
The panel was clear that, while AI allows decisions to be made at scale, both in terms of speed and volume of data considered, poor input data and prompts can lead to incorrect and failed outcomes. The panel discussed steps to mitigate these risks, such as ensuring machine learning models are trained on high-quality data, retraining models for accuracy, and implementing continuous audit and feedback mechanisms. Additionally, the panel considered the steps which could be taken to mitigate the impact of potential issues, such as implementing second-level review by humans for high-impact decisions and establishing appeals processes, where relevant.
5. Opportunities for collaboration
The panel discussed opportunities for collaboration on the use of AI in the product safety world and called for the development of a set of baseline AI/product safety principles, involving industry, regulators, standards organizations, academic experts, and others as the first step to creating a collaborative culture where these new technologies can be tested to further customer safety.
Overall, there was a consensus amongst the panel that AI systems offered a huge opportunity in the product safety space. They emphasised that, to fully reap the benefits of the technology, stakeholder collaboration was critical to foster a collaborative culture where AI technologies can be tested and deployed effectively, with the ultimate goal of furthering product safety.
The Cooley products team will be posting highlights from other plenary panel sessions on our blog throughout the event.
Key contacts:
