TD Bank Group has introduced enterprise-wide Responsible AI Principles, a framework of seven commitments governing how AI is developed, deployed and used across the bank, a governance structure that will also apply to TD Insurance as it expands AI use in claims and customer service.
The principles address compliance with existing financial services laws, transparency and explainability, data use and privacy, fairness, quality and accountability, reliability, and security.
TD is embedding the principles throughout the lifecycle of its AI systems, from design and development through deployment and ongoing monitoring, with every new AI use case required to undergo assessments for explainability, fairness and performance before launch.
The bank is also incorporating the principles into employee education programs to clarify expectations around responsible AI use.
"The future of banking will increasingly be powered by AI," said Sumee Seetharaman, head of artificial intelligence and machine learning practice at TD. Seetharaman said the principles are designed to help the bank scale AI responsibly, with the governance and transparency needed to maintain client trust.
"Trust has always been the foundation of banking, and that won't change as AI becomes more deeply embedded in our industry," said Ajai Bambawale, chief risk officer at TD. Bambawale pointed to colleague-facing virtual assistants as an example already governed by the framework, where controls such as retrieval grounding, human oversight and continuous monitoring help mitigate the risk of inaccurate or inappropriate AI-generated guidance.
The framework is overseen by TD's AI Risk Management team and operationalized by its Trustworthy AI team, a group of scientists focused on the safe and secure use of AI at the bank, supported by governance processes including model risk assessments, privacy impact assessments and independent oversight reviews.
The governance framework lands as TD pushes AI further into claims and customer-facing insurance operations specifically. Speaking on TD's first-quarter 2026 earnings call, chief financial officer Kelvin Tran said the bank expects AI-driven fraud detection, vendor optimization and process redesign to cut insurance claims costs by roughly $150 million over the medium term, part of a broader push toward $1 billion in annual enterprise value from AI that chief executive Raymond Chun first outlined at TD's Investor Day in September 2025. TD has said its AI initiatives generated approximately $170 million in value during 2025.
TD Insurance has separately said it is expanding generative AI use in its contact centers and customer-facing operations. Those deployments will need to operate within the same fairness, explainability and human-oversight requirements the Responsible AI Principles set for the rest of the bank, a governance layer that becomes more significant as AI moves from internal knowledge tools into systems that touch claims decisions and customer communications directly.
TD's move lands amid accelerating AI adoption across Canada's insurance sector more broadly. Finance and insurance was among the sectors with the fastest-growing AI use between mid-2024 and mid-2025, according to Canadian Chamber of Commerce data, with insurers reporting AI use concentrated in data analytics, speech recognition and machine learning.
Manulife ranked third globally among life insurers in the 2026 Evident AI Index and expects to generate more than $1 billion in enterprise value from AI by 2027, while Desjardins Group launched an AI-driven platform in 2025 to personalize insurance products using customer data.
Alongside that adoption, Canadian consumers remain cautious. TD's own research found 61% of Canadians cite inaccurate information as a top concern around AI, with 59% saying they are comfortable with banks using AI only if meaningful human oversight is maintained. KPMG's 2026 insurance trends outlook has flagged Canada's regulatory landscape as entering a more assertive, outcomes-focused period, with insurers under growing pressure to pair AI deployment with stronger data quality and model governance.
TD's principles also arrive ahead of a mandatory deadline that applies directly to insurers.
The Office of the Superintendent of Financial Institutions published the final version of Guideline E-23 on Model Risk Management in September 2025, extending risk management requirements to AI and machine learning models used by all federally regulated financial institutions, including life and property and casualty insurers, not just deposit-taking banks as under the 2017 original.
The guideline takes effect May 1, 2027, after an 18-month transition, and will require documented model inventories, independent validation and explainability standards for AI systems, including those from third-party vendors.
For TD Insurance and other federally regulated insurers, voluntary frameworks like TD's Responsible AI Principles function as an early step toward E-23 compliance rather than a purely reputational exercise.
As other Canadian insurers work through their own E-23 readiness ahead of the 2027 deadline, TD's public framework offers a visible marker of how one federally regulated institution is choosing to get ahead of that timeline.
