Documented cyber programme and certified security module, but little AI governance
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
Marvell lists AI among the Audit Committee's risks and regulates internal AI use in its code of conduct, but discloses no AI responsibility in management, no AI policy and no metrics.
What is documented
- The 2026 proxy statement lists AI among the enterprise risks reviewed by the Audit Committee1
- The CSO leads enterprise risk management and reports quarterly to the Audit Committee3 4 5
- The code of conduct permits only authorised enterprise AI for confidential data and prohibits AI outputs in products or designs7
- The human rights policy contains a commitment to responsible AI2
- The concern line is run by an independent third party, can be used anonymously and is also open to external stakeholders7 3
The case against
No AI governance documented
Sources for Responsibility (10)
- Marvell Technology, Inc. / U.S. SEC: Proxy Statement (DEF 14A) for the annual general meeting of 2026-06-25 · 13 May 2026 · primary source
- Marvell Technology, Inc.: Global Human Rights Policy · 3 November 2026 · secondary source
- Marvell Technology, Inc.: FY24 Sustainability Report · 3 November 2026 · primary source
- Board Cybersecurity (excerpt from 10-K): Marvell Technology, Inc. 10-K Item 1C. Cybersecurity – 2026-03-11 · 11 March 2026 · secondary source
- Marvell Technology, Inc.: Leadership: Derek Hardy · 3 November 2026 · secondary source
- Built In (job listing, Marvell): Director, AI Governance & Enablement · 6 June 2026 · secondary source
- Marvell Technology, Inc.: The Marvell Code – Code of Business Conduct and Ethics · 8 April 2025 · secondary source
- Marvell Technology, Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2026-01-31 · 11 March 2026 · primary source
- Marvell Technology, Inc. (blog): Marvell earns World's Most Ethical Companies designation · 18 March 2026 · secondary source
- SustainabilityReports.com: Marvell Technology Inc 2025 Sustainability Report (metadata) · 1 September 2025 · secondary source
2 · Safety
Marvell documents a cybersecurity programme aligned with ISO 27001 and NIST with third-party testing and an officially validated HSM, but discloses nothing on the security and misuse prevention of its AI chips.
What is documented
- The programme aligned with ISO 27001:2022 and NIST CSF includes a round-the-clock SOC, EDR, DLP and supplier assessment1
- Third parties carry out penetration tests; Internal Audit reviews annually, and tabletop exercises take place1 3
- NIST confirms an active FIPS 140-3 Level 3 validation of the NITROXIII HSM until 20292
- A reporting channel for product vulnerabilities with PGP encryption exists8
- According to an excerpt from the 10-K, there was no material cyber incident in three years3
The case against
Vulnerabilities and reporting channels
- For the QConvergeConsole tool, unsupported since 2022, critical vulnerabilities with CVSS up to 9.4 remained unpatched10 11 12
- The reporting channel guarantees no response times, there is no bug bounty, and the advisories page is incomplete8 9
- LockBit claimed a data theft in 2024; Marvell attributed it to a third-party provider6
AI chips and misuse
- There is no information on testing, end use and diversion of the AI accelerators
- Item 1C of the 10-K was read only via an excerpt, not in the original3
Sources for Safety (15)
- Marvell Technology, Inc.: FY24 Sustainability Report · 3 November 2026 · primary source
- NIST Cryptographic Module Validation Program: Certificate #4700 – NITROXIII CNN35XX-NFBE HSM Family · 30 May 2024 · primary source
- Board Cybersecurity (excerpt from 10-K): Marvell Technology, Inc. 10-K Item 1C. Cybersecurity – 2026-03-11 · 11 March 2026 · secondary source
- Marvell Technology, Inc. / U.S. SEC: Proxy Statement (DEF 14A) for the annual general meeting of 2026-06-25 · 13 May 2026 · primary source
- Marvell Investor Relations: Form 10-K FY2026 (PDF version) · 11 March 2026 · primary source
- The Stack: Marvell data breach · 25 June 2024 · secondary source
- Marvell Technology, Inc.: Marvell, Microsoft and Oracle Establish OCP Working Group to Standardize Data Center Hardware Security · 13 October 2026 · secondary source
- Marvell Technology, Inc.: Product Security Center – Product Security Incident Reporting · 3 November 2026 · primary source
- Marvell Technology, Inc.: Security Advisories · 3 November 2026 · secondary source
- Trend Micro Zero Day Initiative: ZDI-25-733: (0Day) Marvell QConvergeConsole compressConfigFiles Directory Traversal · 31 July 2025 · primary source
- CVEFeed: CVE-2025-6804 – Marvell QConvergeConsole · 7 July 2025 · secondary source
- SentinelOne: CVE-2025-8426: Marvell QConvergeConsole · 31 July 2025 · secondary source
- Marvell Technology, Inc.: The Marvell Code – Code of Business Conduct and Ethics · 8 April 2025 · secondary source
- Marvell Technology, Inc.: Marvell Russia Sanctions Statement · 1 March 2022 · secondary source
- U.S. Senate, Office of Senator Blumenthal (PSI): PSI Majority Staff Report on semiconductor manufacturers' deficient compliance with export controls · 10 September 2024 · secondary source
3 · Transparency
Marvell discloses in the 10-K and the code of conduct that and how it uses AI internally, but publishes no AI report and no information on the tools used or their limitations.
What is documented
The case against
Sources for Transparency (7)
- Marvell Technology, Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2026-01-31 · 11 March 2026 · primary source
- Marvell Technology, Inc. (Kopie via Fortune/Quartr): Form 10-Q for the quarter ended 2026-05-02 · 28 May 2026 · primary source
- Marvell Technology, Inc. / U.S. SEC: Annual Report to Security Holders (ARS) 2026, shareholder letter · 13 May 2026 · secondary source
- Marvell Technology, Inc.: The Marvell Code – Code of Business Conduct and Ethics · 8 April 2025 · secondary source
- Marvell Technology, Inc.: Privacy Policy · 1 January 2025 · secondary source
- Marvell Technology, Inc.: Security Advisories · 3 November 2026 · secondary source
- Marvell Technology, Inc.: Reports and Policies · 3 November 2026 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: within the review period there is no judgment, regulatory decision or fine, and none of the remaining controversies concerns harm caused by the use of AI.
5 · Controversies (2)
- minorIn 2025 the Zero Day Initiative published critical vulnerabilities in the Marvell tool QConvergeConsole for which there is no patch, as the product has been unsupported since 2022. · 31 July 2025 · unklar
- minorThe ransomware group LockBit claimed in 2024 to have stolen Marvell data; Marvell attributed the matter to an incident at a third-party provider. · 25 June 2024 · unklar
6 · What remains open
- AI governance in management: responsibility, bodies and reporting line
- Published AI policy, risk classification and approval procedure
- End-use and customer screening for AI chips and defence applications
- Content of the FY25 Sustainability Report, whose retrieval failed
- Items 1C and 3 of the 10-K in the original, and legal matters in Part II of the 10-Q
- ISO 27001 certificate and its scope
7 · How it was assessed
Cut-off date: 3 November 2026 · 40 searches, 38 fetches.
Primary documents read (11)
- Form 10-K FY2026 (S01 HTML, two retrievals; S02 PDF, two retrievals; Item 1C and 3 not included)
- Item 1C Cybersecurity via extract (S03)
- Proxy Statement DEF 14A 2026 (S04, two retrievals)
- Form 8-K Voting Results 2026 (S05)
- Form 10-Q Q2 FY2027 (S16, two retrievals) and Q1 FY2027 (S17)
- ARS 2026 with shareholder letter (S30)
- Code of Business Conduct and Ethics 2025 (S06)
- FY24 Sustainability Report (S07, two retrievals)
- Global Human Rights Policy (S12), Supplier Code of Conduct 2026 (S29), Privacy Policy (S28)
- Product Security Center and Security Advisories (S13, S14)
- Independent and official findings: NIST CMVP certificate #4700 (S34), ZDI-25-733 (S25), PSI notice (S38)
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0