Independently tested AI for study reports, but little documented AI governance
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
Merck names AI as a board priority and has AI principles from 2023, but discloses no responsible committee, no persons responsible for AI and no approval process.
What is documented
- The board's 2026 priorities include digital technologies, including AI1
- Cybersecurity sits with the Audit Committee, data protection with the Governance Committee1 2
- Seven AI principles from May 2023 refer to internal AI guidelines4
- According to the press, there is a formal policy on generative AI with technical guardrails5
- A company-wide training programme on generative and agentic AI covers responsible use8 9
- The data protection programme is certified under APEC CBPR and BCR and offers an external complaints channel via BBB National Programs10 11
The case against
No documented governance
Missing metrics and audits
- No AI-specific metrics, external audits or complaints channels were found
- The Ethics & Values chapter of the Impact Report remained unreadable and may contain further information12
Sources for Responsibility (12)
- Merck & Co., Inc. (merck.com): 2026 Merck Proxy Statement · 8 April 2026 · primary source
- Merck & Co., Inc. (SEC EDGAR): Definitive Proxy Statement (DEF 14A) 2025 · 9 April 2025 · primary source
- Merck & Co., Inc. (SEC EDGAR): Definitive Proxy Statement (DEF 14A) 2024 · 28 May 2024 · primary source
- Merck Sharp & Dohme (msd.com): Developing and using AI ethically and responsibly · 1 May 2023 · secondary source
- Fortune: How Merck leveled up its cybersecurity strategy since the NotPetya cyberattack that led to a $1.4bn dispute · 22 October 2024 · secondary source
- Dice (job posting by Merck & Co., Inc.): Artificial Intelligence (AI) Ethics and Governance Senior Specialist · 27 May 2026 · secondary source
- Merck & Co., Inc. (merck.com): Merck and Google Cloud Partner to Accelerate Agentic AI Enterprise Transformation · 22 April 2026 · secondary source
- Merck & Co., Inc. (merck.com): 5 ways we're transforming artificial intelligence into impact · 9 January 2026 · secondary source
- Merck & Co., Inc. (merck.com): Merck Expands Innovative Internal Generative AI Solutions Helping to Deliver Medicines to Patients Faster · 25 June 2025 · primary source
- Merck Sharp & Dohme (msdprivacy.com): MSD Privacy – Our Commitment to Privacy (US) · 5 July 2026 · primary source
- Merck & Co., Inc. (merck.com): Purpose for Progress: Merck Impact Report Summary 2025/2026 · 1 August 2026 · primary source
- Merck & Co., Inc. (merck.com): Purpose for Progress: Merck Impact Report 2025/2026 · 1 August 2026 · primary source
2 · Safety
Merck had its AI for study reports independently tested for hallucinations and describes a cyber programme with external assessments and red-teaming, but the test covers only one application.
What is documented
- In the Global AI Assurance Pilot, an independent assessor tested the iRAP application for faithfulness to the source, contradictions and hallucinations1
- The AI drafts of clinical study reports are reviewed by medical writers2 3
- Item 1C describes a CISO-led programme with external assessments, tabletop and red-teaming exercises and supplier assessment5 6
- Cybersecurity is overseen by the Audit Committee; incidents are escalated to an Incident Assessment Team5
- According to the press, Merck relies on zero trust, attack emulation by third parties and guardrails against model training on company data7
The case against
Limited testing coverage
- The independent test covers only one use case and one test dataset and excluded prompt injection, data leakage and bias1
- There is no evidence of validation for AI in manufacturing, trial recruitment, the supply chain and on the agent platform12
- Item 1C was read only via an account, not directly from the SEC filing5
Sources for Safety (12)
- AI Verify Foundation (Singapore) / Resaro: MSD x Resaro – Global AI Assurance Pilot Case Study · 1 May 2025 · primary source
- Merck & Co., Inc. (merck.com): Merck Expands Innovative Internal Generative AI Solutions Helping to Deliver Medicines to Patients Faster · 25 June 2025 · primary source
- McKinsey & Company: With gen AI, Merck and McKinsey transform clinical authoring · 27 June 2025 · secondary source
- Merck Sharp & Dohme (msd.com): Developing and using AI ethically and responsibly · 1 May 2023 · secondary source
- Board Cybersecurity (reproduced from the Merck & Co. 10-K): Merck & Co., Inc. 10-K Item 1C. Cybersecurity – 2026-02-24 · 24 February 2026 · primary source
- Board Cybersecurity (reproduced from the Merck & Co. 10-K): Merck & Co., Inc. 10-K Cybersecurity GRC – 2026-02-24 · 24 February 2026 · primary source
- Fortune: How Merck leveled up its cybersecurity strategy since the NotPetya cyberattack that led to a $1.4bn dispute · 22 October 2024 · secondary source
- Merck & Co., Inc. (merck.com): Purpose for Progress: Merck Impact Report Summary 2025/2026 · 1 August 2026 · primary source
- Levi & Korsinsky, LLP (AccessNewswire): Merck Sharp & Dohme LLC Data Breach under Investigation by Levi & Korsinsky, LLP · 2 December 2025 · secondary source
- Claim Depot: Merck Data Breach: Financial Account Information Exposed · 17 November 2025 · secondary source
- Bloomberg Law: Merck Unit Faces Online Privacy Case That Tests New Legal Theory · 20 November 2023 · secondary source
- Merck & Co., Inc. (merck.com): Merck and Google Cloud Partner to Accelerate Agentic AI Enterprise Transformation · 22 April 2026 · secondary source
3 · Transparency
Merck promises in its principles to disclose AI interactions, but apart from a case study on its study report AI it publishes no system descriptions, model cards or data information.
What is documented
The case against
Implementation not documented
- No labelling is documented for the field force chatbot and AI-generated content for healthcare professionals and patients2
- The privacy page does not mention AI or automated decisions4
- Model and system cards, a transparency report and information on training data are missing; the models of the internal platform are known only from secondary sources7
Sources for Transparency (8)
- Merck Sharp & Dohme (msd.com): Developing and using AI ethically and responsibly · 1 May 2023 · secondary source
- Merck & Co., Inc. (merck.com): 5 ways we're transforming artificial intelligence into impact · 9 January 2026 · secondary source
- Merck & Co., Inc. (merck.com): Purpose for Progress: Merck Impact Report 2025/2026 · 1 August 2026 · primary source
- Merck Sharp & Dohme (msdprivacy.com): MSD Privacy – Our Commitment to Privacy (US) · 5 July 2026 · primary source
- AI Verify Foundation (Singapore) / Resaro: MSD x Resaro – Global AI Assurance Pilot Case Study · 1 May 2025 · primary source
- Fortune: How Merck leveled up its cybersecurity strategy since the NotPetya cyberattack that led to a $1.4bn dispute · 22 October 2024 · secondary source
- IntuitionLabs: GPTeal: Merck's Generative AI Strategy for Pharma R&D · 10 October 2025 · secondary source
- Merck & Co., Inc. / Mayo Clinic (Q4 investor platform): Merck and Mayo Clinic Announce New Research and Development Collaboration to Support AI-Enabled Drug Discovery and Precision Medicine · 18 February 2026 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: no AI-related harm, authority decision or settlement is documented in the review period. The two data protection cases are minor and have no AI reference.
5 · Controversies (2)
- minorA security incident at the relocation service provider Graebel affected personal data of Merck employees in 2025; Merck notified those affected and the authorities. · 22 September 2025 · unklar
- minorIn 2023 a class action was filed against Merck's subsidiary Intervet (HomeAgain), alleging that tracking software on its website is an unlawful “pen register”. · 17 November 2023 · unklar
6 · What remains open
- Explicit AI responsibility at board level and named persons responsible for AI
- Current AI policy, risk classification and approval process, also for purchased and agentic AI
- Validation and security evidence for AI in manufacturing, trial recruitment, sales and on the agent platform
- Content of the Ethics & Values chapter of the Impact Report and of any AI risk factor in the 10-K
- Status of the HomeAgain class action
7 · How it was assessed
Cut-off date: 3 November 2026 · 43 searches, 30 fetches.
Primary documents read (9)
- Form 10-K FY2025 (Item 1, pipeline section; Item 1A/1C/3 truncated)
- Form 10-K FY2025 Item 1C (reproduced via board-cybersecurity.com, two pages)
- Proxy Statements 2026, 2025 and 2024 (board oversight, skills matrix, agenda items)
- Form 10-Q Q2 2026 (up to character 100,000)
- Purpose for Progress Impact Report 2025/2026 (extract) and Summary 2025/2026; MSD Impact Report 2024/2025 (extract)
- AI principles «Developing and using AI ethically and responsibly» (May 2023)
- MSD Privacy (US) and Values and Standards
- Press releases: CSR platform (2025-06-25), Mayo Clinic (2026-02-18), Google Cloud (2026-04-22); article «5 ways» (2026-01-09)
- Case study MSD x Resaro (AI Verify Foundation, Global AI Assurance Pilot)
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0