Clear oversight of cyber risk, but AI governance only in principle and little product information
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
The proxy statement documents audit committee oversight of technology and cyber risk and AI training up to board level, but explicit AI oversight is missing, and AI risk governance is described only in principle.
What is documented
- The Audit Committee has primary oversight of technology and cyber risk; the board received training on generative AI1
- The board considered a separate technology committee with reference to AI risks and rejected it1
- A Chief AI Officer also heads the AI unit Kensho3
- The code of conduct allows only internal AI tools for confidential data and requires a security review before uploading data4
- According to Accenture, around 40,000 employees completed basic AI training including responsible AI content9
The case against
Governance only in principle
No external accountability
- No AI metrics, no AI report and no certification of AI management such as ISO/IEC 42001 were found
- The data protection complaint channel does not cover AI-based outputs10
Sources for Responsibility (10)
- S&P Global Inc. / U.S. SEC: Proxy Statement 2026 (DEF 14A, tm261443-1_def14a.htm) · 31 March 2026 · primary source
- S&P Global Inc.: Annual Report 2025 · 31 December 2025 · secondary source
- PYMNTS: S&P Global AI Chief Eyes 'Exciting' Pace of Agentic AI Innovation · 7 May 2025 · secondary source
- S&P Global Inc.: S&P Global Code of Business Ethics 2026 · 3 November 2026 · secondary source
- S&P Global Inc.: 2025 Impact Report · 31 December 2025 · secondary source
- S&P Global Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2025-12-31 (spgi-20251231.htm) · 31 December 2025 · primary source
- S&P Global Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2024-12-31 (spgi-20241231.htm) · 31 December 2024 · primary source
- Built In NYC (job listing S&P Global Ratings): Generative AI Monitoring Analyst, Compliance & Control, S&P Global Ratings · 8 January 2026 · secondary source
- Accenture: S&P Global empowers talent, builds AI-ready workforce (case study) · 3 November 2026 · secondary source
- S&P Global Inc.: S&P Global Corporate Privacy Policy (Update Effective September 13, 2024) · 13 September 2024 · secondary source
2 · Safety
The 10-K documents penetration testing, independent audits and an incident response plan, but there are only company statements on the validation and error rates of its own AI products.
What is documented
- The 10-K describes an ERM framework, a CISO reporting to the board twice a year and an incident response plan6
- External penetration tests and independent audits review the design and effectiveness of controls6
- AI-supported attacks are named as a risk in the 10-K14
- The S&P AI Benchmarks test financial LLMs with 600 questions and a published methodology1
- The code of conduct requires AI-related security incidents to be reported8
The case against
Testing of AI products not documented
Sources for Safety (15)
- Kensho Technologies (S&P Global): Launched: S&P AI Benchmarks, a solution that evaluates LLMs for finance and business · 3 November 2026 · primary source
- PYMNTS: S&P Global AI Chief Eyes 'Exciting' Pace of Agentic AI Innovation · 7 May 2025 · secondary source
- Kensho Technologies (S&P Global): Grounding Agent – Overview (documentation) · 3 November 2026 · secondary source
- S&P Global Inc. (PR Newswire): S&P Global Introduces CreditCompanion™, Enhancing RatingsDirect® on S&P Capital IQ Pro with Advanced AI Technology · 21 May 2025 · secondary source
- Built In NYC (job listing S&P Global Ratings): Generative AI Monitoring Analyst, Compliance & Control, S&P Global Ratings · 8 January 2026 · secondary source
- S&P Global Inc. / U.S. SEC: 10-K 2025, XBRL report R26 “Cybersecurity Risk Management and Strategy Disclosure” (Item 1C) · 31 December 2025 · primary source
- S&P Global Inc.: 2025 Impact Report · 31 December 2025 · secondary source
- S&P Global Inc.: S&P Global Code of Business Ethics 2026 · 3 November 2026 · secondary source
- JD Supra: CPPA Acts to Enforce Against Data Brokers' Failure to Register · 12 January 2026 · secondary source
- Digital Policy Alert: California Privacy Protection Agency issued USD 62,600 fine against S&P Global for failure to register as data broker · 30 December 2025 · secondary source
- DeXpose: Vect Ransomware Targets Business Services Leader S&PGLOBAL · 16 April 2026 · secondary source
- Rankiteo Blog: Checkmarx, Guesty, LiteLLM and S&P Global: Don't pay VECT a ransom · 15 April 2026 · secondary source
- S&P Global Inc. (PR Newswire): S&P Global Launches Agentic AI-Powered Credit Memo Builder™ to Streamline Credit Analysis · 4 June 2026 · secondary source
- S&P Global Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2025-12-31 (spgi-20251231.htm) · 31 December 2025 · primary source
- S&P Global Inc. (PR Newswire): S&P Global brings Credit Ratings to the S&P Global AI Data Portal · 28 September 2026 · secondary source
3 · Transparency
The 10-K discloses AI risks and product announcements mention source citations, but model and system cards and information on models, training data and error limits are missing for all products.
What is documented
- The 10-K names accuracy and bias risks of AI and the use of AI in internal processes1 2
- The Credit Memo Builder cites data sources in the text and states that it does not replace credit analysis6
- According to the announcement, ChatIQ offers traceable answers, and AI newsletters are labelled “Powered by GenAI”3
The case against
Missing model information
Ratings business without disclosure
- According to the annual report, AI use in the ratings business is increasing, but there are no public rules on it12
Sources for Transparency (12)
- S&P Global Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2025-12-31 (spgi-20251231.htm) · 31 December 2025 · primary source
- S&P Global Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2024-12-31 (spgi-20241231.htm) · 31 December 2024 · primary source
- S&P Global Inc. (PR Newswire): S&P Global Transforms S&P Capital IQ Pro Experience with the Launch of New Generative AI-Powered Capabilities · 12 November 2024 · secondary source
- S&P Global Inc. (PR Newswire): S&P Global Introduces CreditCompanion™, Enhancing RatingsDirect® on S&P Capital IQ Pro with Advanced AI Technology · 21 May 2025 · secondary source
- S&P Global Inc. (PR Newswire): ChatIQ Leads S&P Global's Latest Capital IQ Pro Update with Expanded AI-Powered Research Capabilities · 6 October 2026 · secondary source
- S&P Global Inc. (PR Newswire): S&P Global Launches Agentic AI-Powered Credit Memo Builder™ to Streamline Credit Analysis · 4 June 2026 · secondary source
- Kensho Technologies (S&P Global): Grounding Agent – Overview (documentation) · 3 November 2026 · secondary source
- PYMNTS: S&P Global AI Chief Eyes 'Exciting' Pace of Agentic AI Innovation · 7 May 2025 · secondary source
- Kensho Technologies (S&P Global): Launched: S&P AI Benchmarks, a solution that evaluates LLMs for finance and business · 3 November 2026 · primary source
- Kensho Technologies (S&P Global): Kensho Privacy Policy · 1 July 2023 · secondary source
- S&P Global Inc.: S&P Global Corporate Privacy Policy (Update Effective September 13, 2024) · 13 September 2024 · secondary source
- S&P Global Inc.: Annual Report 2025 · 31 December 2025 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: the CPPA fine concerns a registration obligation and the ransomware claim is unconfirmed; none of the cases concerns harm caused by AI.
5 · Controversies (2)
- minorThe group Vect claims to have stolen 250 GB of internal data from S&P Global; S&P Global has not confirmed this. · 15 April 2026 · unklar
- minorCalifornia's data protection authority CPPA fined S&P Global USD 62,600 for failing to register as a data broker. · 30 December 2025 · Vergleich
6 · What remains open
- Public AI principles, an AI body and AI metrics
- Validation, red-teaming and error rates of the AI products and of AI in the ratings business
- Rules and misuse controls for providing ratings and market data to third-party AI systems
- External audit or certification of AI management
- Model and system cards and the use of customer data for model training
7 · How it was assessed
Cut-off date: 3 November 2026 · 42 searches, 35 fetches.
Primary documents read (7)
- Form 10-K 2025 (sec.gov, main document in two parts up to Item 1A, XBRL R26 Item 1C, R22 and R93 Contingencies)
- Form 10-K 2024 (sec.gov, Item 1A)
- Proxy Statement 2026 (DEF 14A) in two parts, together with DEFA14A
- Form 10-Q Q2 2026 (sec.gov, main document in two parts, XBRL R20 Legal & Regulatory Matters)
- Annual Report 2025, 2025 Impact Report (three parts), Code of Business Ethics 2026
- SEC Press Release 2024-114, ESMA Notice 2023-03-24
- Privacy statements: S&P Global (Sept. 2024) and Kensho (July 2023), Kensho grounding documentation
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0