generated: '2026-08-26'
method: searched
source: https://www.preferred.jp/en/company/aipolicy
name: Preferred Networks trust, compliance and AI governance
description: >-
  Preferred Networks publishes no dedicated trust centre or trust.* subdomain, and
  probe-security-programs.py found neither one nor a vulnerability-disclosure programme. It does,
  however, publish a named certification and three downloadable governance/security policy documents
  from its corporate site, which are recorded here.
trust_center:
  published: false
  probed:
  - url: https://trust.preferred.jp/
    status: DNS NXDOMAIN
  note: No trust portal, subprocessor list, or downloadable audit-report request flow was found.
certifications:
- name: ISO/IEC 27001
  scope: >-
    AI Products and Solutions Division — development, commissioned work, and provision of products
    and services.
  status: certified
  statement: >-
    "AI Products and Solutions Division is ISO 27001-certified for development, commissioned work and
    provision of products and services."
  source: https://www.preferred.jp/en/company/aipolicy
  certificate_number: not published
  auditor: not published
  scope_note: >-
    The certification is stated for one division, not company-wide. PFN does not publish which legal
    entity, which registrar, or whether the PLaMo API platform falls inside the certified scope.
policies:
- name: Information Security Basic Policy
  url: http://preferred.jp/downloads/pfn_information_security_basic_policy_en.pdf
  format: PDF
  language: en
- name: Cloud Security Policy
  url: https://static.preferred.jp/downloads/pfn_cloud_security_policy_en.pdf
  format: PDF
  language: en
- name: Privacy Policy
  url: https://www.preferred.jp/en/policy/
  format: HTML
  language: en
ai_governance:
  published: true
  url: https://www.preferred.jp/en/company/aipolicy
  framework: >-
    An AI governance framework PFN describes as aligned with Japanese government guidelines, under
    management commitment, with published principles of Transparency, Risk-readiness and Integrity.
  safety_evaluation:
    benchmark: HELM Safety (Stanford Center for Research on Foundation Models)
    claim: >-
      PFN states PLaMo 3.0 Prime achieved safety performance at or above comparable overseas models
      across HELM Safety categories, using safety data provided by NICT.
    source: https://www.preferred.jp/ja/news/pr20260622
    independent_verification: none published
data_handling:
  zero_data_retention: >-
    Advertised as a plan feature (ZDR — input data not retained) on the Standard and Provider plans;
    not available on the announced Free plan.
  training_on_customer_data: >-
    "Input data not used for training" is a Standard/Provider plan feature; the Free plan does not
    carry it.
  data_residency: >-
    PFN markets PLaMo as a domestic (Japanese) model; third-party coverage reports requests are
    processed on servers in Japan, but no first-party data-residency commitment was located in the
    documentation, terms page (client-rendered) or product site.
  caveat: >-
    safety_identifier values may be persisted server-side; the API reference instructs callers not to
    put personal or sensitive data in that field.
vulnerability_disclosure:
  published: false
  security_txt: false
  bug_bounty: none found
  probed:
  - url: https://www.preferred.jp/.well-known/security.txt
    status: 404
  - url: https://api.platform.preferredai.jp/.well-known/security.txt
    status: 404
  - url: https://docs.plamo.preferredai.jp/.well-known/security.txt
    status: 404
  note: >-
    No security.txt, no disclosure policy page, and no HackerOne/Bugcrowd/Intigriti programme was
    found. No Security pointer is emitted.