AI oversight and approval procedure documented, but little on the limits of the models
Overall rating “solid”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
Lilly documents oversight, responsible executives and an approval procedure for AI, but no metrics and no external audit.
What is documented
- According to the 2026 proxy statement the board oversees AI and receives regular updates from management1 2 3
- The audit committee oversees AI-related risks, the science and technology committee AI in research4 1
- The procedure «Using AI Responsibly» requires prior approval of new AI applications and risk impact assessments4
- Supplier contracts contain AI quality clauses; approved AI products are on a curated list4
- Since 2024 a Chief AI Officer has been responsible for AI strategy across all functions5
The case against
Implementation not documented
- No metrics on reviewed use cases, training or bias audits
- No external audit or certification of AI management found
- A formal classification of AI risks is not described4
Grievance channels
- For AI concerns there is only the general reporting line «Speak Up at Lilly»9
Sources for Responsibility (10)
- Eli Lilly and Company (SEC EDGAR): Definitive Proxy Statement (DEF 14A) 2026 · 17 March 2026 · primary source
- Eli Lilly and Company (SEC EDGAR): Definitive Proxy Statement (DEF 14A) 2025 · 21 March 2025 · primary source
- Eli Lilly and Company (SEC EDGAR): Definitive Proxy Statement (DEF 14A) 2024 · 22 March 2024 · primary source
- Eli Lilly and Company (sustainability.lilly.com): Use of AI (Impact Areas) · 1 July 2026 · primary source
- Eli Lilly and Company (Investor Relations): Lilly appoints Thomas J. Fuchs as company's first Chief AI Officer · 8 October 2024 · primary source
- Eli Lilly and Company (SEC EDGAR): Form 10-K for the fiscal year ended December 31, 2025 · 12 February 2026 · primary source
- Purdue University, Daniels School of Business: Lilly's AI use cases · 21 January 2025 · secondary source
- Recruiter.com (Eli Lilly job posting): Senior Director of AI Quality Assurance · 3 November 2026 · secondary source
- Eli Lilly and Company (lilly.com): Ethics & Compliance · 3 November 2026 · primary source
- Eli Lilly and Company: Global Privacy Notice – United States · 14 October 2025 · primary source
2 · Safety
The cybersecurity programme in the 10-K is concrete; Lilly describes AI-specific testing only itself.
What is documented
- Item 1C describes a programme based on NIST CSF with external auditors, maturity assessments, threat simulations and an incident response playbook6
- The audit committee meets the CIDO, CISO and Chief Privacy Officer; suppliers are reviewed and contractually bound6 7
- With TuneLab the data stay with the participants; only encrypted model updates are transmitted8
- According to reports the supercomputer runs on premises and disconnected from the network, with logging and a model registry1 2
- The UK data protection authority holds no information on self-reported data breaches by Lilly UK9
The case against
AI testing not evidenced
Sources for Safety (14)
- Eli Lilly and Company (lilly.com): New supercomputer could change future medicine · 26 February 2026 · secondary source
- Constellation Research (Larry Dignan): Eli Lilly: Lessons from running an AI factory · 1 October 2026 · secondary source
- Purdue University, Daniels School of Business: Lilly's AI use cases · 21 January 2025 · secondary source
- Recruiter.com (Eli Lilly job posting): Senior Director of AI Quality Assurance · 3 November 2026 · secondary source
- Yseop: Eli Lilly case study: content automation · 13 June 2023 · secondary source
- Board Cybersecurity (reproduced from Eli Lilly's 10-K): ELI LILLY & Co 10-K Item 1C. Cybersecurity – 2026-02-12 · 12 February 2026 · primary source
- Board Cybersecurity (reproduced from Eli Lilly's 10-K): ELI LILLY & Co 10-K Cybersecurity (Tracker) – 2026-02-12 · 12 February 2026 · primary source
- Revvity Signals: Lilly TuneLab-at-Revvity Federated Learning Program (Information Sheet) · 1 July 2026 · primary source
- Information Commissioner's Office (UK): Disclosure log IC-369751-M1Y5 · 18 March 2025 · primary source
- BioPharma Dive: Eli Lilly offers biotechs AI models via TuneLab · 9 September 2025 · secondary source
- Drug Discovery & Development: Lilly launches TuneLab, sharing AI drug discovery models with biotechs · 11 September 2025 · secondary source
- Eli Lilly and Company (sustainability.lilly.com): Use of AI (Impact Areas) · 1 July 2026 · primary source
- Eli Lilly and Company: Global Privacy Notice – United States · 14 October 2025 · primary source
- Chimicles Schwartz Kriner & Donaldson-Smith LLP: Are Drug Companies and Wellness Companies Sharing Your Health Data? – LillyDirect · 4 December 2025 · secondary source
3 · Transparency
Only the use of AI with personal data is concretely disclosed; model cards and limits are missing.
What is documented
The case against
Sources for Transparency (7)
- Eli Lilly and Company: Global Privacy Notice – United States · 14 October 2025 · primary source
- Eli Lilly and Company (sustainability.lilly.com): Use of AI (Impact Areas) · 1 July 2026 · primary source
- Constellation Research (Larry Dignan): Eli Lilly: Lessons from running an AI factory · 1 October 2026 · secondary source
- Revvity Signals: Lilly TuneLab-at-Revvity Federated Learning Program (Information Sheet) · 1 July 2026 · primary source
- BioPharma Dive: Eli Lilly offers biotechs AI models via TuneLab · 9 September 2025 · secondary source
- Drug Discovery & Development: Lilly launches TuneLab, sharing AI drug discovery models with biotechs · 11 September 2025 · secondary source
- Fierce Biotech: Eli Lilly invites biotechs to use its new AI platform to help develop their own drugs · 9 September 2025 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: no AI-related harm, decision or settlement is documented for the period; the law firm inquiry into LillyDirect is a minor controversy without a documented lawsuit.
5 · Controversies (1)
6 · What remains open
- The AI risk factor in the 2025 10-K and Item 1C in the original were not read.
- How many AI applications were reviewed, with which results, training rates and bias audits?
- Is there an external audit or certification of AI management?
- What validation results and limits do own models and those provided to third parties have?
- The «U.S. Principles on AI» could not be retrieved.
- How will the inquiry into tracking pixels on LillyDirect end?
7 · How it was assessed
Cut-off date: 3 November 2026 · 44 searches, 38 fetches.
Primary documents read (13)
- Form 10-K FY2025 (Forward-Looking Statements, Item 1, part of Item 1A)
- Form 10-K FY2025 Item 1C (reproduced via board-cybersecurity.com)
- Proxy Statements 2026, 2025 and 2024 (AI sections, board oversight, agenda items)
- Form 10-Q Q2 2026 and Q1 2025 (Forward-Looking Statements, financial section)
- Use of AI (impact page, as at 2026-07-01)
- Global Privacy Notice – United States (2025-10-14)
- Ethics & Compliance (Red Book, Speak Up at Lilly)
- Position Statements (AI entry)
- Press releases: Chief AI Officer (2024-10-08) and NVIDIA Supercomputer (2025-10-28)
- TuneLab-at-Revvity Federated Learning Program (July 2026)
- FDA Warning Letter MARCS-CMS 716475
- Complaint: State of Texas ex rel. Health Choice Alliance v. Eli Lilly (2025-08-11)
- ICO Disclosure Log IC-369751-M1Y5
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0