How does Charles Schwab govern AI and protect clients from AI risk?
Charles Schwab governs AI mainly through its model risk framework and through controls on how outside AI can reach client accounts. Its 10-K for 2023 describes a model policy, a centralized model database and model risk ratings without mentioning AI; the 10-K for 2024 adds that Schwab 'manages model risk, including use of artificial intelligence'; and the 10-K for 2025 adds that 'Prior to the use of any artificial intelligence, we employ procedures which require cross-functional review from applicable oversight functions.' The same filing says AI has 'resulted in increased risk of external fraud', including impersonation. For clients, CEO Rick Wurster said on October 15, 2026 that Schwab blocks unauthorized AI agents that try to log in on clients' behalf, lets client LLMs in only through its MCP servers, and does 'not allow them to move money'; on October 28 he said a particular agent tries to log in '30, 40, 50 times' a week. Schwab's Claude agreement with Anthropic, reported in September 2026, bars retaining connector data for model training. Its July 2026 earnings release warns that generative AI output 'may be inaccurate, containing hallucinated, stale, or incomplete information'. Schwab has not published AI principles, and its 2026 proxy statement does not mention AI oversight.[1][2][3][4][5][6][7][8]
Data as of Oct 31, 2026 · Schwab profile updated Oct 19, 2026 · 16 sources cited on this page · 141 banks profiled on this site · Markdown
What AI controls has Schwab disclosed?
| Control | What it does | Disclosed | Source |
|---|---|---|---|
| Model risk framework | A model inventory with a distinct record and risk rating for each model; independent validations, annual reviews and performance monitoring; AI named within it since the 10-K for 2024 | February 2025 | 10-K for 2024[2] |
| Review before any AI use | 'Prior to the use of any artificial intelligence, we employ procedures which require cross-functional review from applicable oversight functions' | February 2026 | 10-K for 2025[3] |
| Unauthorized agents blocked | AI agents that try to log in on behalf of clients are blocked: 'It is not a safe way for our clients to access Schwab, and we are blocking it to protect clients and our firm' | October 15, 2026 | Q3 2026 call[4][9] |
| MCP access with limits | Client LLMs connect through Schwab's MCP servers with secured authentication; 'We do not allow them to move money because we think there is too much risk there to the client and to the firm' | October 15, 2026 | Q3 2026 call[4] |
| No model training on connector data | Under the Anthropic connector agreement 'no data will be retained for model training'; client account numbers are hidden from Anthropic | September 21, 2026 | InvestmentNews[6] |
| Read-only Claude access | Schwab has limited Claude for Financial Advisors to read-only access | October 28, 2026 | InvestmentNews[5] |
| Output warning to clients | Portfolio Insights footnote: 'Generative AI output may be inaccurate, containing hallucinated, stale, or incomplete information' | July 21, 2026 | Q2 2026 earnings release (8-K)[7] |
How has Schwab's 10-K language on AI changed?
The Charles Schwab Corporation's annual reports, filing by filing.
| Filing | What it says about AI |
|---|---|
| 10-K for 2023 (filed February 23, 2024) | Information security risks rise in part because of 'the use of artificial intelligence technologies' by organized crime, hackers and others; the model risk section describes a centralized model database and risk ratings with no reference to AI[1] |
| 10-K for 2024 (filed February 26, 2025) | 'Schwab manages model risk, including use of artificial intelligence, through use of policies, standards, and controls which evaluate the conceptual and technical soundness of models'[2] |
| 10-K for 2025 (filed February 25, 2026) | AI 'has resulted in increased risk of external fraud by enhanced or novel techniques, including those involving impersonation'; adds cross-functional review before any AI use[3] |
How does Schwab handle AI-enabled fraud, and what has it not disclosed?
Schwab's June 2026 client article on AI fraud says it is 'investing in AI-powered detection systems' and that Schwab Voice ID builds an encrypted voiceprint from 'hundreds of elements' of a client's voice; its director of voice products says the systems are updated 'to adapt to emerging threats like deepfakes and synthetic voices'.[10]
At Schwab IMPACT on October 28, 2026, Wurster said 'it doesn't seem that a lot of clients have a lot of interest in giving up their credentials and their financial experience to these agents', and that Schwab is building two approved routes instead: its own assistant and LLM access through an MCP server.[5]
Not on the record: published AI principles, a named AI governance committee, AI incident or complaint data, and board-level AI oversight. The 2026 proxy statement, made available on or about April 6, 2026, does not discuss AI.[8][3]
What does the Schwab profile record on this topic?
The rows below are the entries of the Schwab AI strategy profile that bear on this page, as of Oct 19, 2026; the profile holds the full record of 27 dated moves, 19 use cases and 38 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| Machine-learning fraud detection | Long-running ML models across the business. | In production | [11] |
| Schwab Assistant | Chat and voice AI assistant for clients' most frequent service needs, built with an AI agent firm with human handoffs and guardrails; employee pilot from July 2026. | Pilot | [12][13][14] |
| AI-powered fraud detection against AI-enabled scams | The eCrimes research and detection team is investing in AI systems that identify patterns and unusual behaviour at scale; a published patent application covers fraud detection during authentication. | In production | [10][15] |
| AI agents for internal engineering tasks | Dooher: agents writing code for internal tasks cut project timelines about 20%. | In production | [16] |
| Data-minimised Claude connector | No retention for model training, masked account numbers, no SSNs or dates of birth, under the Anthropic agreement. | In production | [6] |
| AI in the model risk framework | Policies, standards and controls on AI models, cross-functional review before any AI use, a rated model inventory and independent validation. | In production | [3] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 21, 2026 | Anthropic barred from training on Schwab data | Dooher: 'no data will be retained for model training'; account numbers masked; Social Security numbers and dates of birth never passed; terms apply to advisors on Claude tiers below enterprise. | [6] |
| Jun 5, 2026 | Schwab's fraud team: 'we're using AI, too' | Zack Rosebrock, director of eCrimes research and detection, says Schwab is investing in AI-powered detection systems that find patterns at a scale impossible for people, against scams that are more personal and more believable. | [10] |
| Apr 16, 2026 | Q1 call: 'Schwab is already an AI-enabled company' | Wurster lays out an AI roadmap: an investor AI assistant from June built with 'a leading AI agent firm', a voice agent with human handoffs, generative search on schwab.com and Wealth.com AI estate planning, alongside AI tools for all 33,000 employees. | [12] |
| Feb 25, 2026 | 10-K: AI inside model risk management | The annual report says Schwab manages the model risk of AI through policies, standards and controls, requires cross-functional review before any AI use, and flags AI-enabled impersonation fraud. | [3] |
| Nov 10, 2025 | IMPACT 2025: 'the year of the notetaker'; agents write internal code 20% faster | Dooher says notetakers are the top advisor use; Schwab's engineers use AI agents for internal tasks with about 20% faster project completion; Lisa Salvi expects agentic experimentation in 2026 and adoption from 2027. | [16] |
| Feb 26, 2025 | FY2024 10-K: AI enters model-risk language | The first annual report to say Schwab manages the model risk of AI through policies, standards and controls, and to flag AI-enabled external threats. | [2] |
| Feb 5, 2025 | Schwab Knowledge Assistant disclosed | Gen-AI search with cited answers for client-service representatives; ML fraud and NLP models noted. | [11] |
Does Charles Schwab have AI principles?
Schwab has not published a responsible AI principles document. Its governance is described in the 10-K: AI sits inside the model risk framework, with a model inventory, risk ratings, independent validation and monitoring, and since the 10-K for 2025, cross-functional review by oversight functions before any AI is used.
Does Schwab let AI agents access client accounts?
Not unauthorized ones. Wurster said on October 15, 2026 that Schwab blocks agents that try to log in on a client's behalf. Client LLMs can connect through Schwab's MCP servers with Schwab controlling authentication, but they cannot move money, and Claude for Financial Advisors is limited to read-only access.
Does Anthropic train on Schwab data?
No, under the connector agreement. Alison Dooher, head of AI for Schwab Advisor Services, told InvestmentNews in September 2026 that 'no data will be retained for model training', extending that to advisers with Claude plans below the enterprise tier, and that client account numbers are hidden from Anthropic.
What AI risks does Schwab disclose to investors?
Its 10-K filings cite AI used by criminals and hackers, a higher risk of external fraud including impersonation to gain access to client accounts, and AI within model risk. Its July 2026 earnings release warns that generative AI output may be inaccurate, hallucinated, stale or incomplete.
Who oversees AI at Schwab's board?
Schwab has not said. Its 2026 proxy statement does not mention AI, and the 10-K describes oversight functions that review AI before use without naming a committee. The one AI executive Schwab has put on the record in this context is Alison Dooher, head of AI for Schwab Advisor Services.
- Annual Report on Form 10-K for the fiscal year ended December 31, 2023
- Annual Report on Form 10-K for the fiscal year ended December 31, 2024
- Annual Report on Form 10-K for the fiscal year ended December 31, 2025
- Schwab (SCHW) Q3 2026 Earnings Call Transcript
- Schwab CEO says firm is blocking rogue AI agents from client accounts
- Schwab, Anthropic block model training in Claude deal
- Exhibit 99.1: Schwab Reports Record Quarterly Revenue and Earnings (Form 8-K)
- 2026 Proxy Statement (DEF 14A)
- Schwab blocks unauthorized AI agents, points clients to 'Charley'
- The Evolution of AI Fraud: Offense and Defense
- AI on the other end of the line (but you're not talking to a computer)
- Schwab (SCHW) Q1 2026 Earnings Call Transcript
- Schwab (SCHW) Q2 2026 Earnings Call Transcript
- Leading today. Defining tomorrow.
- US 2025/0133081 A1: Detecting fraud during user authentication processes (Charles Schwab & Co.)
- Schwab: 2025 was the 'year of the notetaker,' but AI's next role is less certain
When Schwab 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