Documented IT security structure, but hardly any rules on AI and a data leak
Overall rating “limited”. No deduction for controversies. The superscript numbers link directly to the source.
1 · Responsibility
Advantest documents an independent board majority and compliance reporting, but on AI offers only the formula of “governed” use without a policy or responsibility.
What is documented
- Five of the board's nine directors are independent, and it met thirteen times in fiscal year 20251 2
- Risks are managed through the Executive Management Committee and the Internal Control Committee; information security is a key risk with named risk owners4 3
- Advantest measures training rates for compliance, information security and privacy8 9
- A multilingual, anonymous hotline is also open to outsiders; the board receives reports on it9 2
The case against
No AI governance documented
Missing accountability
- AI-specific training, metrics, ISO/IEC 42001 certifications or external AI audits were not found
- Whether the board dealt with AI or the 2026 cyber incident is not documented
Sources for Responsibility (10)
- Advantest Corporation (Meldung an die Tokyo Stock Exchange): Corporate Governance Report (as of 2026-08-05) · 5 August 2026 · primary source
- Advantest Corporation: Integrated Annual Report 2026 (FY2025) · 3 November 2026 · secondary source
- Advantest Corporation: Integrated Annual Report 2025 (FY2024) · 3 November 2026 · secondary source
- Advantest Corporation: Risk Management System (website) · 3 November 2026 · secondary source
- Advantest Corporation: Annual Securities Report and Business Report, 84th term (English translation) · 26 June 2026 · secondary source
- Advantest Corporation (Ablage Yahoo Finance): Annual Securities Report and Business Report, 84th term · 26 June 2026 · secondary source
- Advantest America (press release): Advantest Pioneers a New Era of AI-Powered Semiconductor Testing · 6 October 2025 · secondary source
- Advantest Corporation: Sustainability Report 2025, Governance chapter · 3 November 2026 · primary source
- Advantest Corporation: Initiatives to Promote Ethical Business Practices (website) · 3 November 2026 · primary source
- Advantest Corporation: The Advantest Way (Corporate philosophy, Ethical Standards) · 3 November 2026 · secondary source
2 · Safety
IT protection is concretely documented, but not specific to AI, and an officially registered ransomware incident with the leak of sensitive personal data weighs against it.
What is documented
- Advantest is certified to ISO/IEC 27001 at six sites, from Japan in 2021 to Korea and India in 20252 3
- Advantest uses multi-factor authentication, a CSIRT, a SOC and external vulnerability testing2 3
- A PSIRT for product security has existed since fiscal year 20264
- After the attack, Advantest isolated systems, brought in Unit 42 and informed law enforcement9 7
- Advantest reported the incident to US attorneys general and offered those affected 18 months of identity monitoring5 7
The case against
2026 ransomware incident
- An unauthorised third party entered the network and stole data including social security, ID, medical and financial information5 7 6
- Those affected were notified only about 233 days after detection; the total number has not been disclosed10 11
- The authority's register gives 23 January 2026 as the date of intrusion, the company 15 February as the date of detection5 6
AI functions not verified
- On testing, validation or monitoring of the ML models in ACS RTDI there are only self-reported statements without a description of procedures
- Usage restrictions or misuse monitoring for AI functions were not found
Sources for Safety (11)
- Advantest America (press release): Advantest Pioneers a New Era of AI-Powered Semiconductor Testing · 6 October 2025 · secondary source
- Advantest Corporation: Information Security (Governance/Compliance website) · 3 November 2026 · primary source
- Advantest Corporation: Sustainability Report 2025, Governance chapter · 3 November 2026 · primary source
- Advantest Corporation: Product Security (website) · 3 November 2026 · primary source
- Office of the Attorney General, State of California: Data Security Breach Report: Advantest America, Inc. · 5 October 2026 · primary source
- Advantest Corporation: Notice regarding cybersecurity incident (Advantest Responds to Cybersecurity Incident) · 19 February 2026 · primary source
- Commonwealth of Massachusetts, Office of Consumer Affairs and Business Regulation (Ablage): 2026-1718 Advantest America Inc. (notification letter) · 6 October 2026 · primary source
- Semiconductor Engineering (sponsored blog, author Michael Chang, Advantest): True Zero Trust Combats IC Manufacturing Security Challenges · 8 February 2024 · secondary source
- Advantest Corporation: IT System Incident Update · 4 March 2026 · primary source
- Tech Insider: Advantest data breach: 233 days from ransomware to notification · 7 October 2026 · secondary source
- BleepingComputer: Advantest confirms personal information stolen in ransomware attack · 7 October 2026 · secondary source
3 · Transparency
Advantest mentions its use of AI in reports, but describes neither how its AI functions work nor their limits or data basis.
What is documented
The case against
Missing disclosure
- The privacy notice contains no statement on automated decisions or AI processing5
- Labelling rules for AI-generated output were not found
- A system description for ACS RTDI with models, training data and data separation between customers is missing
Sources for Transparency (5)
- Advantest Corporation: Integrated Annual Report 2026 (FY2025) · 3 November 2026 · secondary source
- Advantest Corporation: Annual Securities Report and Business Report, 84th term (English translation) · 26 June 2026 · secondary source
- Advantest Corporation (Ablage Yahoo Finance): Annual Securities Report and Business Report, 84th term · 26 June 2026 · secondary source
- Advantest America (press release): Advantest Pioneers a New Era of AI-Powered Semiconductor Testing · 6 October 2025 · secondary source
- Advantest Corporation: Privacy Notice (ver. September 16, 2026) · 16 September 2026 · secondary source
4 · Why the harm rule does not apply
The harm rule does not apply: there is no judgment, regulatory decision or fine on the 2026 cyber incident, and there is no AI link. No pattern of serious AI incidents was found.
5 · Controversies (1)
6 · What remains open
- AI-specific responsibility in the board or in executive management
- Content of the “governed” use of AI: rules, approvals, tools used and providers
- Validation, monitoring and protection of the ML functions in ACS RTDI and separation of customer data
- Cause, extent and total number of people affected by the 2026 data leak; notification to Japanese authorities
- The risk chapter of the securities report for fiscal year 2025 on cyber and AI risks was not readable
7 · How it was assessed
Cut-off date: 3 November 2026 · 41 searches, 34 fetches.
Primary documents read (11)
- Integrated Annual Report 2026 (pp. 2–44, two retrievals)
- Integrated Annual Report 2025 (to approx. p. 37)
- Annual Securities Report and Business Report 84th term, English (to p. 32)
- 有価証券報告書兼事業報告書 第84期, Japanese (to p. 20)
- コーポレート・ガバナンス報告書 dated 2026-08-05 (TSE)
- Sustainability Report 2025, Governance chapter (pp. 116–140)
- 臨時報告書 dated 2026-07-31 (EDINET)
- Voting results and Q&A from the 84th Annual General Meeting
- Website pages: Information Security, Product Security, Ethical Business Practices, Advantest Way, Risk Management System, Privacy Notice
- Corporate announcements regarding the cyber incident of 2026-02-19 (Japanese) and 2026-03-04
- Registry filing with the California Attorney General and notification letter in the Massachusetts filing
Research of 3 November 2026 under Globalance AI Footprint rulebook version 0