In an era where AI capabilities are accelerating faster than governance frameworks can keep pace, TD has taken a grounded, pragmatic approach to building trust into its AI initiatives, setting an industry benchmark. TD’s initiative stands out for its substance, ambition, and execution.
At the heart of its efforts is TD’s Layer 6, the Bank’s in-house AI research and development arm. Far from adopting a single-threaded approach, Layer 6 has built a multi-dimensional framework for trustworthy AI. This is not just box-ticking ethics, but a serious engineering and operational commitment that embeds privacy-enhancing technologies, fairness constraints, anomaly detection, and explainability directly into model development and deployment processes.
TD’s approach is striking for its operational realism. Rather than assuming that AI systems can or should always be the final decision-makers, TD has built human-in-the-loop mechanisms that allow models to defer to humans when confidence thresholds are not met. This is where responsible AI moves beyond principle into practice – recognizing the need for judgement.
The Bank’s 22 patent filings in 2024 related to trustworthy AI underscore the depth of innovation underpinning this work. These are intellectual property investments aimed at aligning AI performance with TD’s values, safeguarding customers, and mitigating emerging risks.
Crucially, TD is treating trustworthy AI as a strategic lever to tackle the governance challenges posed by generative AI and foundation models. In doing so, TD is defining what best practice should look like.
