Own AI leadership and cyber protection documented, but little disclosure to clients
Overall rating “solid”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
UBS documents AI leadership, an AI framework and the embedding of AI governance in three lines of defence, but an explicit AI responsibility at board level and external AI audits are missing.
What is documented
- The 2025 sustainability report names a Chief AI Officer appointed in 2025 as well as an AI framework and an AI policy1 11
- The response to the FSB describes AI controls in all three lines of defence and the shift from a use-case to a system approach7
- The Risk Committee reviews non-financial risks; the Group COO is responsible for group-wide data governance2
- Whistleblowing is possible anonymously; Group Investigations reports quarterly to the board and executive board1 2
- A general complaints channel with a response deadline and reference to the banking ombudsman exists8
The case against
Oversight and rules
Sources for Responsibility (12)
- UBS Group AG (Form 6-K, SEC): UBS Group Sustainability Report 2025 · 9 March 2026 · primary source
- UBS Group AG (SEC, Exhibit 1.2 to Form 20-F): Organization Regulations of UBS Group AG · 9 March 2026 · primary source
- finews.asia: UBS Names First Chief AI Officer to Accelerate Transformation · 16 October 2025 · secondary source
- UBS Group AG: Artificial intelligence (company website) · 1 October 2026 · secondary source
- UBS Group AG: Innovation and AI (company website) · 3 November 2026 · secondary source
- Economist Enterprise (supported by Infosys): How UBS is rewiring for the age of AI · 30 June 2026 · secondary source
- UBS AG (Financial Stability Board, consultation response): Response to the FSB consultation on Sound Practices for Responsible Adoption of AI · 22 July 2026 · primary source
- UBS Switzerland AG: Feedback and complaints – Complaint · 3 November 2026 · primary source
- UBS Group AG: Learning and development (company website) · 3 November 2026 · secondary source
- BankTrack (Kopie des UBS-Berichts): UBS Group Sustainability Report 2025 (copy) · 9 March 2026 · primary source
- UBS Group AG: Sustainability Report 2025 (sustainability-report-2025.pdf) · 9 March 2026 · primary source
- UBS Group AG (SEC): Form 20-F FY2025 (ubs-20251231.htm) · 9 March 2026 · primary source
2 · Safety
The 20-F documents a group-wide security programme with red-team exercises and reports to the board, but test and validation results for generative AI and agents are missing.
What is documented
- A monthly Cyber and Information Security Committee, a Group CISO and quarterly metrics for the board are documented6
- The 20-F names penetration tests, red-team exercises, third-party monitoring and a 24/7 cyber operations centre6
- Models for risk quantification are independently validated1
- According to the response to the FSB, AI controls are adjusted after tests, incidents and near misses2
- According to the CEO, external AI models may not carry out actions for clients3
The case against
AI testing not evidenced
Sources for Safety (12)
- UBS Group AG (SEC, XBRL): Form 20-F FY2025, R52: MD&A – Risk management and control · 9 March 2026 · primary source
- UBS AG (Financial Stability Board, consultation response): Response to the FSB consultation on Sound Practices for Responsible Adoption of AI · 22 July 2026 · primary source
- capital.com (Reuters): UBS says it does not permit the use of external AI models for client transactions · 27 October 2026 · secondary source
- Economist Enterprise (supported by Infosys): How UBS is rewiring for the age of AI · 30 June 2026 · secondary source
- MLex (Law360): Embedding AI in banking comes with robust controls, UBS says · 30 May 2025 · secondary source
- UBS Group AG (SEC, XBRL): Form 20-F FY2025, R54: Cybersecurity Risk Management and Strategy Disclosures · 9 March 2026 · primary source
- UBS Group AG (SEC, Exhibit 1.2 to Form 20-F): Organization Regulations of UBS Group AG · 9 March 2026 · primary source
- UBS Group AG: Privacy notice (legal information, privacy) · 1 October 2022 · primary source
- The Star (Reuters): UBS and Pictet report data leak after cyber attack on provider, client data unaffected · 18 June 2025 · secondary source
- finews.ch: Data of UBS employees ends up on the dark web · 18 June 2025 · primary source
- UBS (USA): Fraud prevention · 3 November 2026 · secondary source
- swissinfo.ch: UBS deploys AI analyst clones · 20 May 2025 · secondary source
3 · Transparency
UBS publishes use cases with figures, but information on limits, client information on the use of AI and AI statements in the privacy notice are missing.
What is documented
- The AI website names principles such as human autonomy, fairness and transparency, as well as use cases with figures2 5
- According to a report, videos with AI avatars of analysts carry the Seal «made with AI»1
- According to UBS, the transaction monitoring system gives analysts auditable explanations for every decision5
The case against
Little information for clients
- The privacy notice (as of 2022) contains no statement on AI, profiling or automated individual decisions4
- No client information on the use of AI in advice, communication and transaction monitoring was found
- Information on limits and error rates of the systems used is missing
Self-description instead of a report
- The information on use cases comes from self-descriptions; a transparency report or AI inventory with risk classes is missing2
Sources for Transparency (7)
- swissinfo.ch: UBS deploys AI analyst clones · 20 May 2025 · secondary source
- UBS Group AG: Artificial intelligence (company website) · 1 October 2026 · secondary source
- BankTrack (Kopie des UBS-Berichts): UBS Group Sustainability Report 2025 (copy) · 9 March 2026 · primary source
- UBS Group AG: Privacy notice (legal information, privacy) · 1 October 2022 · primary source
- UBS Group AG: Innovation and AI (company website) · 3 November 2026 · secondary source
- capital.com (Reuters): UBS says it does not permit the use of external AI models for client transactions · 27 October 2026 · secondary source
- Finextra: UBS underpins AI development with data transformation initiative · 1 December 2024 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: within the window there is no ruling, no regulatory decision and no pattern of serious incidents involving the company's own AI systems. A 2025 fine in Hong Kong concerns a rule-based procedure without AI.
5 · Controversies (1)
6 · What remains open
- The section on responsible use of AI in the 2025 sustainability report (p. 66) could not be read in any copy
- Risk and governance chapters of the 2025 annual report, including non-financial risk, model risk and cybersecurity governance
- Responsibility of the board or a committee for AI risks
- Testing, validation and red-teaming results for generative and agentic AI and governance of employee agents
- Client information and explanations on the use of AI in advice and transaction monitoring
- Remedial measures after the data breach at the supplier Chain IQ
7 · How it was assessed
Cut-off date: 3 November 2026 · 41 searches, 41 fetches.
Primary documents read (8)
- UBS Group Sustainability Report 2025 (SEC 6-K, with read-more; ubs.com PDF up to p. 32; BankTrack copy)
- UBS Group Sustainability Report 2024 (SEC 6-K, part)
- Form 20-F FY2025: main document (cross-references), XBRL R31 Litigation, R52 Risk management (with read-more), R54 Cybersecurity (with read-more)
- Organization Regulations of UBS Group AG, effective 2026-01-05 (Exhibit 1.2)
- Annual Report UBS AG 2025 (up to p. 22)
- UBS Group third quarter 2026 report (6-K) and press release dated 2026-10-28
- UBS consultation response to the FSB on Responsible Adoption of AI (2026-07-22)
- UBS websites: Artificial intelligence, Innovation and AI, Privacy Policy, Complaints Switzerland, Fraud prevention, Learning and development
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0