Skip to content
GLOBALANCE INTELLIGENCEThe economy ahead.
This Week Index 28/100 Sign in
Money & Investments › GI 100 Index
The Great Value ShiftInvestment ThemesGI 100 IndexGI PortfolioStart Future Check
AI FOOTPRINTsolidASSESSMENT 2026GLOBALANCEINTELLIGENCE
Globalance AI Footprint

Assessment as of November 2026 · next assessment 2027.

Result of the assessment by the Globalance research team under the published methodology. Not a certification or quality label.

How we assess ›

How this rating is builtVideo · 0:52
GI 100 Index · Profile

ARM Holdings PLC

United KingdomEnablersAssignment FullData quality Medium

Key messagesolid AI Footprint

Arm designs the computing building blocks on which a large share of artificial intelligence in phones, cars and data centres runs, and has an established process for reporting and fixing security flaws with oversight by the board, even though flaws in graphics drivers have already been exploited by attackers. For the responsible use of artificial intelligence, Arm refers to an internal policy and a dedicated office, but does not disclose their content or effect.

Opportunities

  • Safetysolid

    In the 20-F, Arm documents a NIST-aligned cybersecurity programme with a CISO, annually tested contingency plans, external maturity assessments and oversight by the Audit Committee (S02), as well as a PSIRT process with CNA status, coordinated disclosure and a public bug bounty programme (S11, S13).

    In the 20-F, Arm documents a NIST-aligned cybersecurity programme with a CISO, annually tested contingency plans, external maturity assessments and oversight by the Audit Committee (S02), as well as a PSIRT process with CNA status, coordinated disclosure and a public bug bounty programme (S11, S13). Formal verification of the Confidential Compute Architecture is documented and integrated into the development chain (S16). Against this stand repeated vulnerabilities in the Mali GPU drivers, including CVE-2024-4610 with exploitation in the wild confirmed by Arm, and the circumvention of MTE (S26, S27, S28, S29). AI-specific testing, safety evidence for AI accelerators and misuse safeguards for the AGI CPU are not documented, and functional safety rests on company statements without a named certification body (S14, S15, S34).

  • Transparencysolid

    Arm discloses security flaws through bulletins with CVE identifiers and describes its cyber strategy and governance in concrete terms in the 20-F (S11, S02).

    Arm discloses security flaws through bulletins with CVE identifiers and describes its cyber strategy and governance in concrete terms in the 20-F (S11, S02). For the models provided in the Arm AI Portal, Arm documents origin, licence, evaluation, accuracy loss and the limits of training data information (S18). Transparency remains narrow: the AI Policy and the AI Office's mandate are not published, there is no AI report, and the sustainability report is not externally assured (S08, S07).

Risks

  • Responsibilitylimited

    A general risk and control structure is documented: according to its charter, the Audit Committee is responsible for technology and information security and whistleblowing procedures, a Risk Review Committee reports quarterly, and a Raise a Concern Helpline available around the clock is also open …

    A general risk and control structure is documented: according to its charter, the Audit Committee is responsible for technology and information security and whistleblowing procedures, a Risk Review Committee reports quarterly, and a Raise a Concern Helpline available around the clock is also open to outsiders (S09, S06, S08). For AI itself, however, there are only principles: the code of conduct refers to an unpublished AI Policy and an “AI Office”, the sustainability report names a policy on internal use of AI without content, and neither the board nor its committees have a designated responsibility for AI (S08, S07, S10, S32). Figures, external audits or approval procedures for own and purchased AI (such as the use of OpenAI models in Metis) are not documented (S17, S07).

  • Risk area

    Security of hardware, drivers and firmware in billions of devices that run AI

  • Risk area

    Functional safety of AI features in vehicles built on Arm IP

  • Risk area

    Export control and end use of powerful AI processors

  • No serious controversy recorded.

What lies behind the assessment

Research of 3 November 2026
42searches
36pages and documents read
9primary documents
3controversies checked
5open points, named openly
The full research: pillars, evidence, controversies and open points
AI Footprint · Enabler · As of 3 November 2026
ARM Holdings PLCOverall rating: solid

Arm documents a NIST-aligned security programme with a response team, bug bounty and Audit Committee oversight. For AI, Arm refers to a policy and an “AI Office” but does not disclose their content.

Responsibility

40 %
limitedsolidstrong

General risk oversight is documented, but the AI policy and AI Office only in name.

  • +Audit Committee responsible for technology and information security Source ↗
  • +Raise a Concern Helpline around the clock, also for outsiders Source ↗
  • –AI Policy and AI Office documented in name only Source ↗
  • –No AI responsibility of the board or its committees Source ↗
  • –No approval review described for OpenAI models in Metis Source ↗
All sources for this pillar (9)

Safety

40 %
limitedsolidstrong

NIST-aligned security programme with a response team, but exploited flaws in graphics drivers.

  • +NIST-aligned cyber programme with a CISO and Audit Committee oversight Source ↗
  • +PSIRT with CNA status and a public bug bounty programme Source ↗
  • +Formal verification of the Confidential Compute Architecture Source ↗
  • –Mali driver flaw CVE-2024-4610 exploited in the wild Source ↗
  • –No misuse safeguards or end-use checks for the AGI CPU Source ↗
All sources for this pillar (16)

Transparency

20 %
limitedsolidstrong

Security flaws and models in the AI Portal are disclosed, but the AI Policy is not.

  • +Security bulletins with CVE identifiers and coordinated disclosure Source ↗
  • +Model pages state origin, licence, evaluation and accuracy loss Source ↗
  • –AI Policy and AI Office mandate not published Source ↗
  • –No AI report; sustainability report not externally assured Source ↗
All sources for this pillar (7)
What is missing for “strong”

These steps are in the company’s own hands.

1Responsibility
Disclose the AI Policy and AI Office

Publish the AI Policy, state the AI Office's mandate and staffing and assign responsibility for AI at board level.

2Safety
Document fixes and AI testing

Document the remediation status of the Mali fixes and disclose testing and end-use rules for AI accelerators and the AGI CPU.

3Transparency
Publish an AI report

Publish an AI report with figures and document the limitations of AI features in vehicle IP.

Context: 101 companies assessed

Distribution of overall ratings, no ranking
strong
none yet
solid
← ARM Holdings PLC is one of 35
limited

No company reaches “strong” today. Disclosure on how companies handle AI is still young across the market.

Controversies
3 1 medium, 2 minor

No deduction for controversies.

Largest risk area
Exploited flaws in graphics drivers

Arm confirmed that the Mali driver flaw CVE-2024-4610 was exploited in the wild; whether the fixes reach device makers is only partly documented.

How thoroughly assessed
32 Evidence

9 primary documents read, 42 searches.

Read the full reasoning and all sources
ARM Holdings PLC · Full reasoning

An established security process, but AI policy and AI office in name only

Overall rating “solid”. No deduction for controversies. The superscript numbers link directly to the source.

1 · Responsibility

limitedWeight 40%

Arm documents a general risk and control structure, but for AI there are only principles, with no published policy, no AI responsibility at board level and no figures.

What is documented
  • According to its charter, the Audit Committee is responsible for data protection, technology and information security and whistleblowing procedures1
  • A management Risk Review Committee reports quarterly; the board discusses the use of AI within Arm2
  • The Raise a Concern Helpline is available around the clock, also to people outside Arm; retaliation is prohibited3 6
  • Employees receive annual ethics training and an AI e-learning course, without figures6
The case against
AI rules in name only
  • The code of conduct refers to an unpublished AI Policy and an “AI Office” without a mandate or staffing3
  • The policy on internal use of AI is named without content; the 2019 Trust Manifesto contains no implementation mechanisms6 7
  • Neither the board nor its committees have a designated responsibility for AI4 5
No figures or audits
  • Figures, external audits and approval procedures for own and purchased AI are not documented; the sustainability report is not externally assured6 8
Sources for Responsibility (9)
  1. Arm Holdings plc: Audit Committee Charter · 1 August 2023 · primary source
  2. Arm Holdings plc / SEC EDGAR (Form 6-K): Annual Report and Consolidated Financial Statements for the year ended 31 March 2026 · 31 March 2026 · primary source
  3. Arm Limited: Arm Code of Conduct FYE25 · 1 February 2025 · secondary source
  4. Arm Holdings plc: Governance – Overview & Documentation · 3 November 2026 · secondary source
  5. Arm Holdings plc / SEC EDGAR (Form 6-K): Notice to ADS Holders 2026 Annual General Meeting · 6 August 2026 · secondary source
  6. Arm: A Future Built on Arm: FYE25 Sustainable Business Report · 31 March 2025 · secondary source
  7. Arm: Arm AI Trust Manifesto · 6 November 2019 · secondary source
  8. Arm Newsroom (Mark Hambleton): Arm Metis: agentic AI security framework · 28 May 2026 · secondary source
  9. Arm Holdings plc / SEC EDGAR (Form 6-K): Results of 2026 Annual General Meeting · 9 September 2026 · primary source

2 · Safety

solidWeight 40%

Arm documents a NIST-aligned security programme, a PSIRT with bug bounty and formal verification, but Mali driver flaws were exploited, and AI-specific testing is missing.

What is documented
  • The 20-F describes a NIST-aligned programme with a CISO, annually tested contingency plans, external maturity assessments and Audit Committee oversight6
  • The PSIRT acknowledges reports within two working days, is a CNA and publishes bulletins7
  • A bug bounty programme with safe harbour covers Mali drivers and firmware8
  • The Realm Management Monitor was formally verified; verification is built into the development pipeline, and a bug found was fixed1
  • Arm commits to export, customs and sanctions law and has a trade compliance function14
The case against
Vulnerabilities in own systems
  • Arm confirmed exploitation of CVE-2024-4610 in the Mali drivers; distribution of the fix depends on device makers9 10
  • Further flaws affect the Mali drivers (CVE-2025-0072) and the circumvention of the Memory Tagging Extension (TikTag)11 12
  • The group D1R claims a data theft from arm.com; there is no statement from Arm13
AI-specific testing missing
  • Safety evidence for AI accelerators and misuse safeguards for the AGI CPU are not documented16
  • Certification to ISO 26262 is documented only as a company statement without a certification body or date3 4
Sources for Safety (16)
  1. arXiv (Wu, Xiong, Manino, Stockwell, Cordeiro; Arm Centre of Excellence, University of Manchester): Verifying components of Arm Confidential Computing Architecture with ESBMC · 6 June 2024 · primary source
  2. Arm: Arm's Systematic Approach to Product Security · 3 November 2026 · secondary source
  3. Arm: Embedded Assurance and Certifications · 3 November 2026 · secondary source
  4. Arm: Zena Compute Subsystem – Accelerating Time to Market for Safe, AI-Ready Automotive SoC · 3 November 2026 · secondary source
  5. Arm Newsroom (Mark Hambleton): Arm Metis: agentic AI security framework · 28 May 2026 · secondary source
  6. Arm Holdings plc / SEC EDGAR: Form 20-F FY2026, XBRL report R33: Cybersecurity Risk Management and Strategy Disclosure (Item 16K) · 31 March 2026 · primary source
  7. Arm: Product Security Incident Response · 3 November 2026 · primary source
  8. Intigriti: Arm Bug Bounty Program – Detail · 1 June 2025 · primary source
  9. The Hacker News: Arm Warns of Actively Exploited Zero-Day Vulnerability in Mali GPU Drivers · 11 June 2024 · secondary source
  10. Cyber Security News: Arm Mali GPU Kernel Driver 0-Day Vulnerability Actively Exploited in the Wild · 3 February 2025 · secondary source
  11. cvelogic: CVE-2025-0072 – Mali GPU Kernel Driver allows improper GPU memory processing operations · 2 May 2025 · secondary source
  12. The Register: Arm Memory Tag Extensions leak (TikTag) · 18 June 2024 · secondary source
  13. DeXpose: D1R Ransomware Attack on ARM: Compromise of UK Tech Giant · 13 July 2026 · secondary source
  14. Arm Limited: Arm Code of Conduct FYE25 · 1 February 2025 · secondary source
  15. Arm Holdings plc / SEC EDGAR: Form 20-F for the fiscal year ended March 31, 2026 · 31 March 2026 · secondary source
  16. Tom's Hardware: Arm to sell its new AGI CPU in China · 1 April 2026 · secondary source

3 · Transparency

solidWeight 20%

Arm discloses security flaws with CVE identifiers and documents the models in its AI Portal, but the AI Policy and the AI Office's mandate are not published, and there is no AI report.

What is documented
  • Arm discloses security flaws through bulletins with CVE identifiers and publicly flagged the exploited Mali flaw1 3
  • The 20-F describes cyber strategy and governance in concrete terms2
  • For optimised third-party models in the AI Portal, origin, licence, evaluation, accuracy versus the base model and liability notices are documented5
  • For TikTag, Arm named the affected processor cores6
The case against
AI governance not disclosed
  • The AI Policy and the AI Office's mandate are not published; no AI or transparency report was found4
  • The sustainability report is not externally assured7
Gaps on limitations and overview
  • Limitations and failure modes of AI features in vehicle IP are not documented
  • There is no overview of all models provided and their documentation status
Sources for Transparency (7)
  1. Arm: Product Security Incident Response · 3 November 2026 · primary source
  2. Arm Holdings plc / SEC EDGAR: Form 20-F FY2026, XBRL report R33: Cybersecurity Risk Management and Strategy Disclosure (Item 16K) · 31 March 2026 · primary source
  3. Cyber Security News: Arm Mali GPU Kernel Driver 0-Day Vulnerability Actively Exploited in the Wild · 3 February 2025 · secondary source
  4. Arm Limited: Arm Code of Conduct FYE25 · 1 February 2025 · secondary source
  5. Arm Developer (Arm AI Portal): YOLOv5s INT8 (XNNPACK, ExecuTorch) – Purpose of this Release · 3 November 2026 · primary source
  6. The Register: Arm Memory Tag Extensions leak (TikTag) · 18 June 2024 · secondary source
  7. Arm: A Future Built on Arm: FYE25 Sustainable Business Report · 31 March 2025 · secondary source

4 · Why the harm rule does not apply

The harm rule does not apply: there is no judgment, authority decision or fine concerning harm from AI; the exploited Mali flaw is a driver bug with no link to AI.

5 · Controversies (3)

6 · What remains open

  • Responsibility for AI on the board or in executive management
  • Content and implementation of the AI Policy and the AI Office's mandate
  • End-use checks for the AGI CPU and AI IP
  • Remediation status and distribution of the Mali driver fixes among device makers
  • Confirmation or refutation of the group D1R's data theft claim

7 · How it was assessed

Cut-off date: 3 November 2026 · 42 searches, 36 fetches.

Primary documents read (9)
  • Form 20-F FY2026 (S01, two retrievals up to Item 3.D) as well as XBRL reports R33 Cybersecurity (S02) and R27 Contingencies (S03), FilingSummary
  • Form 20-F FY2025 (S05, Part 1)
  • UK Annual Report and Consolidated Financial Statements FY2026 (S06, two retrievals)
  • AGM 2026: Notice (S32) and Voting Results (S31)
  • Sustainable Business Report FYE25 (S07)
  • Code of Conduct FYE25 (S08) and website (S37)
  • Audit Committee Charter (S09), governance page (S10)
  • PSIRT page (S11), product security page (S12), bug bounty programme (S13), certifications page (S14), Zena CSS (S15)
  • Model page Arm AI Portal (S18), Trust Manifesto 2019 (S19)

Research of 3 November 2026 under Globalance AI Footprint rulebook version 0

We assess documented practice in dealing with AI; this is not an investment recommendation. Three pillars, weights 40/40/20, harm rule.

Research of 3 November 2026 under Globalance AI Footprint rulebook version 0. Documented observations from public sources, no investment recommendation.

Important Notice. This material is provided for information purposes only and does not constitute an offer or solicitation to buy or sell any financial instrument, nor does it constitute investment advice or a recommendation. The GI 100 Index is a rules-based strategy index of Globalance Bank. Investments involve risk, including the loss of capital invested. The final product documents (including FinSA) shall prevail in all cases.

Your company? Information on inclusion in the GI 100 Index →

What comes next