A fixed procedure for security flaws, but hardly any rules of its own for AI
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
Apart from AI being listed among the risks the board discusses and one AI rule in the code of ethics, no AI governance by Broadcom was found.
What is documented
- The board discusses risks with management, including on cybersecurity and AI; the Audit Committee is responsible for cybersecurity, data protection and IT1
- The code of ethics (rev. September 2026) allows confidential, client or personal data to be entered only into expressly authorised generative AI tools3
- An anonymous ethics hotline is also open to third parties; retaliation is prohibited, and contact with the compliance officer and the Audit Committee is provided for3
- According to an excerpt from the 10-K, employees receive annual training on data protection and security5
The case against
No AI governance documented
- An AI committee, a named person responsible for AI or a procedure for the board's handling of AI are not described; the audit charter dates from 20221 2
- No AI policy, risk classification or approval review for its own and purchased AI was found
- The AI ethics principles of the former VMware date from before the acquisition and were not carried over to Broadcom7
Reports missing or not read
- AI reports, AI figures and external AI audits are missing; the Corporate Responsibility Report could not be read8
Sources for Responsibility (8)
- Broadcom Inc. / U.S. SEC: Proxy Statement DEF 14A 2026 · 20 April 2026 · primary source
- Broadcom Investor Relations: Corporate Governance · 3 November 2026 · secondary source
- Broadcom Inc.: Code of Ethics and Business Conduct (Rev. September 2026) · 3 November 2026 · secondary source
- Broadcom Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2025-11-02 · 18 December 2025 · primary source
- Board Cybersecurity (extract from 10-K): Broadcom Inc. 10-K Item 1C. Cybersecurity – 2024-12-20 · 20 December 2024 · secondary source
- Broadcom Inc. / U.S. SEC: Form 8-K, voting results of the 2025 annual general meeting · 21 April 2025 · primary source
- CSRwire (VMware): Ethics in the Age of Generative AI: A Closer Look at the Ethical Principles for VMware's AI · 22 September 2023 · secondary source
- GreenTechLead: Broadcom Cuts Emissions 40% as Renewable Energy and AI Efficiency Drive Sustainability Goals · 28 May 2026 · secondary source
2 · Safety
Broadcom documents cyber oversight with quarterly reports and a disclosure procedure for VMware, but actively exploited zero-day flaws and delayed patches weigh on this pillar.
What is documented
- The board and the Audit Committee oversee cybersecurity; the CIO and CISO report quarterly4
- For VMware there is a coordinated disclosure procedure with deadlines, severity levels, safe harbour and information on active exploitation5
- According to ThaiCERT, patches are available for the actively exploited vCenter flaw CVE-2024-370798
- According to excerpts from the 10-K, the cyber programme follows NIST CSF and ISO 27001, with penetration tests and an annual tabletop exercise; no material incidents are reported3 1
The case against
Exploited flaws in VMware
- Three ESXi flaws were exploited as zero-days (VMSA-2025-0004), one of them now used for ransomware according to reports6 7
- According to NVISO, CVE-2025-41244 was exploited for almost a year; the first advisory did not mention the active exploitation10
- Since late May 2025, customers with perpetual licences without a valid support contract have received security patches only with delay11
Misuse barely addressed
- Information on customer screening, export control and diversion protection is missing; VMware's AI agent control is documented only as a product claim13
Sources for Safety (13)
- Board Cybersecurity (extract from 10-K): Broadcom Inc. 10-K Item 1C. Cybersecurity – 2024-12-20 · 20 December 2024 · secondary source
- Broadcom (security.com): Using AI to Predict, Stop and Recover from the Next Ransomware Incident · 15 April 2025 · secondary source
- Board Cybersecurity (extract from 10-K): Broadcom Inc. 10-K Item 1C. Cybersecurity – 2025-12-18 · 18 December 2025 · secondary source
- Broadcom Inc. / U.S. SEC: Proxy Statement DEF 14A 2026 · 20 April 2026 · primary source
- VMware (Broadcom): VMware External Vulnerability Response and Remediation Policy · 3 November 2026 · primary source
- Tenable: CVE-2025-22224, CVE-2025-22225, CVE-2025-22226: Zero-Day Vulnerabilities in VMware ESXi, Workstation and Fusion Exploited · 4 March 2025 · secondary source
- Security Affairs: CVE-2025-22225 in VMware ESXi now used in active ransomware attacks · 4 February 2026 · secondary source
- ThaiCERT: CISA Adds VMware vCenter Vulnerability (CVE-2024-37079) to KEV Catalog After Active Exploitation · 26 January 2026 · primary source
- UpGuard: Broadcom data breach 2026-01-23 · 26 January 2026 · secondary source
- SecurityBrief UK: Broadcom patches VMware zero-day exploited for nearly a year · 2 October 2025 · secondary source
- The Register: VMware prevents some perpetual license holders from downloading patches · 23 July 2025 · secondary source
- The Register: Broadcom employee data stolen by ransomware crooks following hit on payroll provider · 16 May 2025 · secondary source
- SiliconANGLE: Broadcom's private AI cloud spans infrastructure, agents, data and security · 31 August 2026 · secondary source
3 · Transparency
Broadcom names AI-related business risks and a contract for AI accelerators, but not how its AI functions work or where their limits lie; system or model cards are missing.
What is documented
- The 10-Q mentions a long-term contract for custom AI accelerators without naming the customer2
- The 10-K names AI mainly as a business and cyber risk1
- VMware commits to stating active exploitation in security advisories3
- The AI function Incident Prediction shows probabilities for each predicted step and leaves the decision to analysts5
The case against
Disclosure concerns finances, not AI
No information on the AI functions
- For Incident Prediction, training data are described only roughly; limits and error rates are not stated5
- No model or system cards and no transparency report were found
Sources for Transparency (5)
- Broadcom Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2025-11-02 · 18 December 2025 · primary source
- Broadcom Inc. / U.S. SEC: Form 10-Q for the quarter ended 2026-05-03 · 3 November 2026 · primary source
- VMware (Broadcom): VMware External Vulnerability Response and Remediation Policy · 3 November 2026 · primary source
- SecurityBrief UK: Broadcom patches VMware zero-day exploited for nearly a year · 2 October 2025 · secondary source
- Broadcom (security.com): Using AI to Predict, Stop and Recover from the Next Ransomware Incident · 15 April 2025 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: there is neither an authority decision nor a fine against Broadcom over the exploited security flaws or the data theft at a service provider, and no harm caused by AI is documented.
5 · Controversies (4)
- mediumAccording to NVISO, the VMware flaw CVE-2025-41244 was exploited for almost a year; Broadcom patched it but did not mention the exploitation in the first advisory. · 2 October 2025 · unklar
- mediumThree ESXi flaws were exploited as zero-days, one for ransomware according to CISA; in 2026 the vCenter flaw CVE-2024-37079 was also reported as actively exploited. · 4 March 2025 · unklar
- minorSince late May 2025, customers with a VMware perpetual licence without a valid support contract have received security patches only with delay. · 23 July 2025 · unklar
- minorIn a ransomware attack on a partner of the payroll provider ADP, ID, health and account data of Broadcom employees were stolen. · 16 May 2025 · unklar
6 · What remains open
- Responsibility of the board and management for AI risks
- AI policy and review procedures for AI functions in VMware and Symantec products
- Content of the Corporate Responsibility Report on AI, data protection and product security
- Export control and anti-diversion programme for semiconductors
- Item 1C and Legal Proceedings of the 10-K in the original
- Timing of patches for customers without a support contract
7 · How it was assessed
Cut-off date: 3 November 2026 · 41 searches, 38 fetches.
Primary documents read (9)
- Form 10-K FY2025 (S06, two retrievals; Item 1C and 3 not included)
- Proxy Statement DEF 14A 2026 (S09, two retrievals)
- Form 8-K voting results 2025 (S10)
- Form 10-Q quarter to 2026-05-03 (S36, two retrievals; Legal Proceedings not included)
- Code of Ethics and Business Conduct Rev. Sept. 2026 (S11)
- Corporate governance page (S12)
- VMware External Vulnerability Response and Remediation Policy (S13)
- Earnings call transcript Q3 FY2026 (S02)
- Official and independent findings: ThaiCERT on CVE-2024-37079 (S16), BHRRC response summary (S29)
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0