How does Wells Fargo govern AI and manage AI risk?
Wells Fargo governs AI through its board and its existing risk functions rather than a published AI charter. The 2026 proxy gives the board's Risk Committee 'primary oversight responsibility for technology risk, which encompasses new and emerging technologies such as artificial intelligence', alongside data and model risk; at management level the Model Risk function inside Independent Risk Management validates and monitors models and reports to the chief risk officer. The stated tenets come from CIO Chintan Mehta in September 2026: be 'fully transparent', protect customers' data and money, give the best advice and 'don't cause harm', enforced through adverse-effect bias checks and independent reviews at several development points. The design rule is to keep models away from customer data: Fargo sends only scrubbed, tokenized text to Google's Gemini Flash 2.0 to find intent, and none of the bank's APIs pass through the model. Explainability has a longer record: the model risk team under Agus Sudjianto presented interpretable neural-network methods for lending in 2021 and open-sourced the PiML toolbox. The 10-K has warned since its 2023 edition that AI outputs may be 'deficient or biased'. The regulatory backdrop is the Federal Reserve's 2018 growth cap, imposed over governance and risk-management failures and lifted on June 3, 2025 with the order's other provisions still in place; none of those actions concerned AI. From November 24, 2026 its online agreement says requests sent through third-party AI agents may be treated as the customer's own.[1][2][3][4][5][6][7][8][9][10]
Data as of Oct 31, 2026 · Wells Fargo profile updated Oct 21, 2026 · 21 sources cited on this page · 141 banks profiled on this site · Markdown
Who oversees AI at Wells Fargo, from the board down?
From the proxy statements and the annual report's risk-management section. 'Disclosed' means it is in an SEC filing; 'Stated' means a named executive said it.
| Body or role | What it does for AI | Status | Source date |
|---|---|---|---|
| Risk Committee (board) | 'Primary oversight responsibility for technology risk, which encompasses new and emerging technologies such as artificial intelligence', plus data management risk and model risk; oversees model risk policy, model performance, remediation and model risk appetite metrics | Disclosed | March 18, 2026; February 2026[1][2] |
| Full board | Discusses AI 'on a regular basis': periodic management reviews of market developments and the bank's approach, line-of-business presentations on how AI affects each business, and reports from the Risk Committee chair | Disclosed | March 18, 2026[1] |
| Model Risk function (Independent Risk Management) | Oversight of model risk, with governance, validation and monitoring across the company; reports to the chief risk officer and to the board Risk Committee | Disclosed | February 2024; February 2026[7][2] |
| Head of Technology | Bridget Engle's 2024 scorecard credits her with developing a 'framework to establish the proper infrastructure, training, and governance to oversee expansion of AI capabilities' | Disclosed | March 19, 2025[11] |
| AI lead | Saul Van Beurden, named November 20, 2025, accountable to the CEO for AI progress across the company | Disclosed | November 20, 2025[12][13] |
| Central AI team (hub and spoke) | Faraz Shafiq, head of AI products and solutions: 'Governance, security, controls, prioritization – all of that is centrally controlled'; business teams own the applications; risk conversations happen at the outset | Stated | October 20, 2026[14] |
| Independent product reviews | Independent groups review AI-based products 'at multiple development points' before customers see them; adverse-effect checks compare models to find bias | Stated | September 22, 2026[3] |
What has Wells Fargo published or disclosed on responsible AI, and when?
| Date | Item | What it says |
|---|---|---|
| April 13, 2021 | Explainable AI for lending risk (NVIDIA GTC) | Agus Sudjianto, then EVP and head of Corporate Model Risk: ReLU networks 'can be decomposed and represented exactly into linear sub-models'; the LIFE method generates decline-reason codes from 40 to 80 variables[5] |
| April 4, 2023 | Stanford HAI corporate affiliate program | Joins the program, which the bank describes as a founding membership; more than 4,000 employees had taken part by September 2026[15][3] |
| February 20, 2024 | 2023 annual report risk factor | 'The algorithms or datasets underlying our artificial intelligence may be inaccurate or include other weaknesses that could result in deficient or biased data outputs'[7] |
| June 20, 2024 (reported) | PiML toolbox | Python Interpretable Machine Learning: Sudjianto championed its adoption at Wells Fargo and released it under Apache 2.0, per H2O.ai's announcement of his move there[6] |
| April 8, 2025 | Fargo privacy pipeline | Speech transcribed locally, text scrubbed and tokenized with a small PII-detection model, only then sent to Gemini Flash 2.0 for intent; Mehta: 'We're the filters in front and behind'[4] |
| February 24, 2026 | 2025 10-K risk factor | Adds that 'the complexity and limited transparency of certain AI models can make it challenging to understand why they are generating particular outputs'[2] |
| March 9, 2026 | Head of AI's principles | Van Beurden: continuous learning, simple messaging, relentless execution; return on investment sets priority[16] |
| August 12, 2026 | Agentic AI Foundation | Gold member; Mehta cites 'strong governance, transparency, and accountability'[17] |
| September 22, 2026 | How Wells Fargo builds responsible AI | Four practices: eliminating bias, transparency, offering non-AI alternatives, research partnerships; 'You never let an AI neural network reach conclusions without knowing why'[3] |
| October 13, 2026 | Q3 2026 call | Scharf: the bank will evaluate third-party models and agents for clients 'while at the same time considering the potential risk these tools can introduce'[18] |
| November 24, 2026 (effective) | Online Access Agreement update | A third-party AI tool or agent 'is not operated by Wells Fargo'; requests through it 'may be treated as if they came directly from you'; the bank may restrict or block certain AI tools[10] |
How do Wells Fargo's regulatory orders relate to its AI governance?
On February 2, 2018 the Federal Reserve capped Wells Fargo's total assets at their end-2017 level after finding that the firm 'did not have an effective firm-wide risk management framework in place that covered all key risks', and required better board oversight and risk management. On June 3, 2025 the Fed said Wells Fargo had met the conditions, including third-party reviews, and lifted the cap; the order's other provisions stay in place until terminated.[8][9]
Neither order mentions AI, and Wells Fargo has not been the subject of an AI-specific enforcement action. The link is the machinery: the board oversight and firm-wide risk framework built under the order are what the 2026 proxy now extends to AI as part of technology risk.[8][1]
The most direct supervisory contact on AI so far was about cyber risk: in April 2026 Treasury Secretary Bessent and Fed Chair Powell briefed bank CEOs, including Wells Fargo's, on Anthropic's Mythos model; head of technology Bridget Engle later said it 'truly did find new, novel techniques' for exploitation.[19][20]
What does the Wells Fargo profile record on this topic?
The rows below are the entries of the Wells Fargo AI strategy profile that bear on this page, as of Oct 21, 2026; the profile holds the full record of 44 dated moves, 17 use cases and 60 sources.
| Date | Move | Detail | Source |
|---|---|---|---|
| Aug 12, 2026 | Joins the Agentic AI Foundation as a Gold member | The neutral home of MCP and other agent standards; CIO Chintan Mehta cites interoperability and governance. | [17] |
| Apr 29, 2026 | Wall Street expansion post asset cap | Freed from the Fed cap in 2025, the bank invests in trading and crosses $2 trillion of assets. | [21] |
| Apr 9, 2026 | Wells Fargo CEO briefed on Anthropic Mythos cyber risk | Bessent and Powell convened bank CEOs at Treasury; Bloomberg, via Reuters, listed the Wells Fargo CEO among those present. Wells Fargo declined to comment. | [19] |
| Feb 24, 2026 | 2025 10-K spells out AI model risk | AI informs or automates business decisions, risk management and models; biased or inaccurate data and limited transparency named as risks. | [2] |
Does Wells Fargo have responsible AI principles?
It has stated tenets rather than a formal published policy. CIO Chintan Mehta said in September 2026 that responsible AI means being 'fully transparent', protecting customers' data and money, giving the best advice and not causing harm, and the bank describes four practices: eliminating bias, building in transparency, offering customers non-AI alternatives and research partnerships such as Stanford HAI. Head of AI Saul Van Beurden's three principles (continuous learning, simple messaging, relentless execution) are about adoption, not ethics.
Who is responsible for AI risk at Wells Fargo?
At board level, the Risk Committee has primary oversight of technology risk including AI, plus data and model risk, per the 2026 proxy. In management, the Model Risk function in Independent Risk Management validates and monitors models and reports to the chief risk officer. Saul Van Beurden is head of AI, accountable to CEO Charlie Scharf; Faraz Shafiq's central team controls governance, security and prioritization for AI products.
How does Wells Fargo keep customer data out of AI models?
For Fargo, by architecture: speech is transcribed locally, text is scrubbed and tokenized by an internal small language model that detects personal data, and only then is Google's Gemini Flash 2.0 called to identify intent and entities. Computation, detokenization and API calls stay on the bank's side; Mehta said none of its APIs pass through the LLM.
Does the asset cap have anything to do with AI?
Not directly. The Federal Reserve imposed the cap on February 2, 2018 over consumer abuses and a firm-wide risk management framework that did not cover all key risks, and lifted it on June 3, 2025. Neither order mentions AI. The board oversight and risk framework rebuilt under the order are, however, the structure the bank now uses to oversee AI.
What does Wells Fargo say about customers using AI agents?
Its Online Access Agreement update, effective November 24, 2026, says a third-party AI tool or agent is not operated by Wells Fargo, that information shared with it falls under the provider's terms, that requests through it may be treated as coming directly from the customer, and that the bank may restrict or block certain tools. Scharf said on the October 13, 2026 call that the bank is weighing the risks of third-party models and agents.
What has Wells Fargo published on explainable AI?
Its model risk group was an early publisher. In April 2021 Agus Sudjianto, then head of Corporate Model Risk, presented at NVIDIA GTC on decomposing ReLU neural networks into linear sub-models and on the LIFE method for decline-reason codes. He championed the PiML interpretable-ML toolbox at Wells Fargo, which was released under Apache 2.0; he left for H2O.ai in 2024.
- Wells Fargo & Company 2026 Proxy Statement (DEF 14A)
- Wells Fargo & Company 2025 Annual Report (Form 10-K, Exhibit 13): risk factors
- How Wells Fargo builds responsible artificial intelligence
- Wells Fargo's AI assistant just crossed 245 million interactions – no human handoffs, no sensitive data exposed
- XAI Explained at GTC: Wells Fargo Examines Explainable AI for Modeling Lending Risk
- H2O.ai Taps Agus Sudjianto to Make GenAI Trustworthy (former head of Wells Fargo model risk; PiML)
- Wells Fargo & Company 2023 Annual Report (Form 10-K Exhibit 13): risk factors and risk management
- Federal Reserve restricts Wells Fargo's growth until firm improves governance and controls
- Federal Reserve announces Wells Fargo is no longer subject to the asset growth restriction from the Board's 2018 enforcement action
- We Are Updating Our Online Access Agreement (effective November 24, 2026; page retrieved 2026-10-30)
- Wells Fargo & Company 2025 Proxy Statement (DEF 14A)
- Wells Fargo Names Saul Van Beurden to Lead Artificial Intelligence
- Wells Fargo Names Saul Van Beurden to Lead Artificial Intelligence; Kleber Santos to Expand Responsibilities and Serve as Co-CEO of Consumer Banking and Lending with Van Beurden
- 4 keys to Wells Fargo's strategy to transform the bank with AI
- Wells Fargo Joins Stanford HAI Corporate Affiliate Program
- Wells Fargo is in 'growth mode.' The bank's AI head thinks three principles will get it where it wants to go.
- Agentic AI Foundation Welcomes 57 New Members, Gaining Major Financial Services Players and APAC Leaders
- Wells Fargo (WFC) Q3 2026 earnings call transcript
- Bessent, Powell warned bank CEOs about Anthropic model risks, sources say
- The Most Powerful Women in Banking, #11, Bridget Engle, Wells Fargo
- Wells Fargo's Long-Anticipated Wall Street Expansion Is Here
When Wells Fargo moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning