These criteria allow an organization to show its business partners that it is doing everything it reasonably can. OneNym.id publishes in full what it checks during an assessment and what must be met for the AI & Cloud Security Assessment Certificate to be issued. Business partners who receive the Certificate can confirm what it means on this page.
11
areas
40
Required
34
Recommended
17
of them on generative AI
SCOPE
Purpose and scope
These criteria apply to organizations that use cloud services, SaaS, and generative AI in their business (mainly Microsoft 365, Google Workspace, Azure, AWS, Google Cloud, Claude, ChatGPT, and Copilot). The assessment covers configuration, operating rules, records, and organizational arrangements as of the assessment date. Malware analysis of PCs or servers, penetration testing, physical security, and determinations of legal compliance are out of scope.
These criteria are not a certification scheme such as ISO/IEC 27001 (ISMS) or the PrivacyMark; they are assessment criteria defined independently by OneNym.id. When the government's security assessment scheme for supply chains, currently under development, comes into effect, a mapping between that scheme and each item will be added.
The two documents issued
CERTIFICATE
AI & Cloud Security Assessment Certificate
A one-page document for presenting to business partners. It states the grade, scope, expiry date, and the result for each area, with the result for generative AI use shown on a separate line. It does not describe weaknesses in detail.
REPORT
AI & Cloud Security Assessment Report
A report covering every item, held only by the client. It states the result for each item, the basis for that result, and improvement priorities.
RESULTS
Result categories
Each item is either Required or Recommended, and its result is one of the following four. Where there is no evidence, the result is Fail; a verbal explanation alone is never sufficient for a Pass.
Pass
The requirement is met and has been confirmed by evidence
Conditional pass
Part of the requirement is not yet met, but the impact is limited and a time-bound improvement plan has been agreed
Fail
The requirement is not met, or cannot be confirmed for lack of evidence
Not applicable
The relevant service or business process is not used (the reason is stated in the Report)
GRADES
Grades
The Certificate is issued at one of three grades according to scope and level of achievement. Stars are not used, to avoid confusion with the ★ ratings of the government's assessment scheme.
STANDARD
Standard
Cloud, SaaS, and generative AI (the 10 areas other than END)
No Fail on any of the 37 Required items outside END
PLUS
Plus
The Standard scope plus business endpoints (END)
The Standard conditions are met, and there is no Fail on any of the 3 Required END items
ADVANCED
Advanced
Same as Plus
All 40 Required items are Pass (with no Conditional pass), and at least 24 of the 34 Recommended items are Pass
ISSUANCE
Issuance, validity, and revocation
No more than 3 Required items may be a Conditional pass, and the improvements and supporting evidence must be submitted within 90 days of the issue date.
The Certificate is valid for one year from the assessment completion date. It can be renewed by completing a reassessment before it expires.
OneNym.id does not publish the names of companies to which it has issued a Certificate. There is no public list; the client decides to whom it shows its Certificate.
Each Certificate is assigned a non-guessable ID and a verification code. Only a person who enters both on the verification page is shown the company name, grade, expiry date, and status (valid, expired, or revoked).
The Certificate is revoked if conditions remain unresolved after the improvement deadline, if submitted materials prove to be inaccurate, if the Certificate has been altered, or if it is presented as covering more than its stated scope.
If a serious incident affecting the assessed scope comes to light, the client is required to report it, and the relevant areas are re-examined.
How items are verified
Assessments are, in principle, carried out with read-only permissions, and each item is verified with one of the following three types of evidence. Any content organized with the help of AI is checked against the evidence by an assessor before a result is determined.
Configuration
Admin consoles, configuration exports, and output from verification scripts
Records
Policies, registers, procedures, records of training and exercises, and restore test results
Interview
Confirmation of actual practice. An item is rated Pass on interview alone only where no record could exist
CRITERIA
Assessment items
There are 74 items across 11 areas (40 Required, 34 Recommended). Generative AI, with 17 items, is the most heavily weighted area. "ISO ref." gives the numbers of related controls in ISO/IEC 27001:2022 Annex A, provided for reference only as an aid to understanding. It does not indicate conformity with ISO.
GOVGovernance and policy
ID
Requirement
Level
Evidence
ISO ref.
GOV-1
An information security lead has been appointed, and this is known across the organization
Required
Records · Interview
5.2
GOV-2
An inventory of the cloud services, SaaS, and generative AI used in the business exists, and an administrator is assigned to each
Required
Records
5.9, 5.23
GOV-3
An information security policy is documented and reviewed at least once a year
Required
Records
5.1
GOV-4
The security of contractors and partners that handle personal information is checked before a contract is concluded
Recommended
Records
5.19, 5.20
GOV-5
Rules for remote work and for business use of personal devices are documented and communicated
Required
Records
6.7
GOV-6
At least once a year, information assets and threats are identified and control priorities are reviewed (risk assessment)
Recommended
Records
Clause 6.1
IAMIdentity and access
ID
Requirement
Level
Evidence
ISO ref.
IAM-1
Multi-factor authentication is required for all users
Required
Configuration
8.5
IAM-2
There are 2 to 5 top-level administrators, and they use dedicated administrator accounts separate from their everyday accounts
Required
Configuration
8.2
IAM-3
Legacy authentication (basic authentication and connection methods that do not support multi-factor authentication) is disabled
Required
Configuration
8.5
IAM-4
Accounts of departing employees are disabled within 1 business day of their departure date (checked against recent departures)
Required
Configuration · Records
5.18
IAM-5
Shared accounts are not used. Where one is unavoidable, its owner, purpose, and multi-factor authentication are managed
Required
Configuration · Records
5.16
IAM-6
Administrators use phishing-resistant authentication (passkeys or FIDO2 security keys)
Recommended
Configuration
8.5
IAM-7
Access is restricted to managed devices or approved locations, for example through Conditional Access
Recommended
Configuration
8.1, 8.5
IAM-8
An emergency administrator account exists, and an alert is raised whenever it is used
Recommended
Configuration
8.2
IAM-9
External user (guest) accounts are reviewed at least every six months, and those no longer needed are removed
Required
Configuration · Records
5.18
IAM-10
Each user's permissions are reviewed at least once a year, and permissions no longer needed are removed when a user changes roles
Recommended
Records
5.18
IAM-11
A password manager is provided, and password reuse is prevented
Recommended
Configuration · Interview
5.17
DATData sharing and personal information
ID
Requirement
Level
Evidence
ISO ref.
DAT-1
Sharing that allows anyone with the link to open content is disabled or has an expiry date set
Required
Configuration
5.14
DAT-2
Automatic forwarding of email to external addresses is blocked
Required
Configuration
5.14
DAT-3
A register records where customer and employee personal information is held and who can access it
Required
Records
5.12, 5.34
DAT-4
External sharing is reviewed at least twice a year
Recommended
Configuration · Records
5.14
DAT-5
Sensitivity labels or data loss prevention (DLP) rules are in use
Recommended
Configuration
5.12, 8.12
DAT-6
Files are exchanged with external parties by methods that allow access to be controlled, such as sharing links or file transfer services, rather than by emailing password-protected ZIP files
Recommended
Records · Interview
5.14
DAT-7
Retention periods for personal information are defined, and information past its retention period is deleted
Recommended
Records
5.34, 8.10
MAILEmail and spoofing protection
ID
Requirement
Level
Evidence
ISO ref.
MAIL-1
SPF, DKIM, and DMARC are configured for the organization's domains, and reports on spoofed email are received
Required
Configuration
5.14
MAIL-2
Scanning of attachments and links in inbound email is enabled
Required
Configuration
8.7
MAIL-3
The DMARC policy is set to quarantine or stricter
Recommended
Configuration
5.14
MAIL-4
Email from outside the organization is marked as external, and a rule requires requests to change payment details to be verified through a separate channel
Recommended
Configuration · Records
5.14, 6.3
AIGenerative AI use
ID
Requirement
Level
Evidence
ISO ref.
AI-1
There is a list of AI tools approved for business use, and a person is designated to approve new tools
Required
Records
5.23
AI-2
Rules on what information may and may not be entered into AI are documented and communicated
Required
Records · Interview
5.10
AI-3
Approved tools are covered by contracts or settings that prevent input from being used for training
Required
Configuration · Records
5.23
AI-4
AI tools are managed through organizational accounts (SSO or an admin console), and access is disabled when an employee leaves. Business use of personal accounts is prohibited
Required
Configuration · Records
5.16, 5.18
AI-5
Each AI API key has a designated owner and the minimum necessary permissions, is not hard-coded, and is reissued at least once a year
Required
Configuration · Records
5.17, 8.2
AI-6
Usage logs (audit logs) for AI tools are retained and can be investigated in the event of an incident
Recommended
Configuration
8.15
AI-7
The scope of operations that AI agents may perform automatically, and the operations that require human approval, are defined
Recommended
Records · Configuration
8.2
AI-8
Features for sharing AI conversations or projects outside the organization (such as public links) are disabled or require approval
Required
Configuration
5.14
AI-9
Data sources connected to AI (connectors, MCP servers, plugins) require approval and are managed in an inventory
Required
Configuration · Records
5.15, 5.23
AI-10
The incident response plan covers AI-related incidents (confidential information entered by mistake, leaked API keys, erroneous actions by agents)
Required
Records
5.24
AI-11
The retention period and storage location (region) of AI tool conversation histories and uploaded files are known and defined
Recommended
Configuration · Records
5.33, 8.10
AI-12
Credentials used by AI agents are dedicated identities separate from those of people, with the minimum necessary permissions
Recommended
Configuration
5.16, 8.2
AI-13
AI that handles external input (email, web pages, inquiries) has restricted access to confidential data and restricted permissions to take actions (prompt injection protection)
Recommended
Configuration · Records
8.26
AI-14
AI-written code is reviewed by a person before it goes into production, and the extent to which AI coding tools can access secrets and production environments is defined
Recommended
Records · Configuration
8.25, 8.28
AI-15
Use of unapproved AI services, browser extensions, and desktop apps can be identified and, where necessary, restricted (shadow AI protection)
Recommended
Configuration
5.9, 8.23
AI-16
A rule requires a person to review AI output before it is used in customer communications or important business decisions
Recommended
Records · Interview
5.10
AI-17
Contracts with AI vendors are checked for data protection, subcontractors, and incident notification
Recommended
Records
5.19, 5.20
APPConnected apps
ID
Requirement
Level
Evidence
ISO ref.
APP-1
The scope within which ordinary users can consent to connected apps is restricted, and all other consents are approved by an administrator
Required
Configuration
5.15
APP-2
Connected apps with broad permissions are reviewed at least once a year
Required
Configuration · Records
5.18
APP-3
There are no integrations that are unused or have no known owner
Recommended
Configuration
5.18
APP-4
Administrators are notified of consent to new integrations and of expanded permissions for existing integrations
Recommended
Configuration
8.16
ENDEndpoints (Plus and above)
ID
Requirement
Level
Evidence
ISO ref.
END-1
Operating systems and browsers on business endpoints update automatically, and no operating system that is out of support is in use
Required
Configuration · Records
8.8
END-2
Antivirus or EDR is enabled on all endpoints
Required
Configuration
8.7
END-3
Storage on laptops and smartphones is encrypted
Required
Configuration
8.1, 8.24
END-4
The status above is monitored centrally through mobile device management (MDM)
Recommended
Configuration
8.1
END-5
Lost devices can be remotely locked and wiped
Recommended
Configuration
8.1
END-6
Ordinary users are not given administrator rights on their devices
Recommended
Configuration
8.2
LOGLogging and detection
ID
Requirement
Level
Evidence
ISO ref.
LOG-1
Audit logs for key services (Microsoft 365, Google Workspace, and cloud activity logs) are enabled
Required
Configuration
8.15
LOG-2
Audit logs and sign-in records can be accessed for at least 180 days (stored externally where the service's retention period is shorter)
Required
Configuration
8.15
LOG-3
The recipients of critical security alerts and the people responsible for reviewing them are designated
Required
Configuration · Interview
8.16
LOG-4
Logs are retained for at least one year, and logs from multiple services can be analyzed together
Recommended
Configuration
8.15, 8.16
LOG-5
Suspicious activity, such as sign-ins from abroad, impossible travel, and mass downloads, is detected and alerted on
Recommended
Configuration
8.16
BCKBackup and recovery
ID
Requirement
Level
Evidence
ISO ref.
BCK-1
The scope and method of backup for important data (including data held in SaaS) are defined
Required
Records · Configuration
8.13
BCK-2
A restore from backup is tested at least once a year, and the results are recorded
Required
Records
8.13
BCK-3
Backups are stored where they cannot be deleted or altered using everyday administrator permissions
Recommended
Configuration
8.13
BCK-4
For each critical business process, the time within which it must resume and the point to which data must be recoverable (recovery objectives) are defined
Recommended
Records
5.30
IRIncident response
ID
Requirement
Level
Evidence
ISO ref.
IR-1
A documented response plan exists that sets out contacts, decision-makers, and procedures for the first 24 hours
Required
Records
5.24, 5.26
IR-2
Procedures for reporting a personal information breach (preliminary and final reports to Japan's Personal Information Protection Commission (PPC), and notification of affected individuals) are defined
Required
Records
5.24, 5.31
IR-3
A procedure exists for exporting and preserving logs in the event of an incident, and the responsible staff are able to carry it out
Required
Records · Interview
5.28
IR-4
A tabletop exercise is held at least once a year, and the results are reflected in the plan
Recommended
Records
5.24, 5.27
IR-5
External contacts for incidents (lawyers, cyber insurer, forensic investigators, support contacts for services in use, police consultation services, key business partners) are listed
Required
Records
5.5, 5.6
PPLPeople and training
ID
Requirement
Level
Evidence
ISO ref.
PPL-1
Security training (including phishing and AI usage rules) is given to all staff at least once a year, and attendance is recorded
Required
Records
6.3
PPL-2
A point of contact for reporting suspicious email or events is designated and communicated
Required
Records · Interview
6.8
PPL-3
A phishing simulation is conducted at least once a year
Recommended
Records
6.3
PPL-4
Security and AI usage rules are explained to new joiners (including agency and contract staff)
Recommended
Records
6.3
PPL-5
Confidentiality undertakings are obtained from employees and contractors
Required
Records
6.6
SUPPORT
Support in meeting the criteria
OneNym.id does more than assess. It also supports the client in drafting documents such as policies, AI usage rules, and incident response plans, improving configurations, training staff, and running tabletop exercises. Support ranges from providing templates to doing the work on the client's behalf. Documents produced with this support are rated Pass only where there is evidence that the client has approved them and actually operates by them, and every item that received support is identified in the AI & Cloud Security Assessment Report.
The AI & Cloud Security Assessment Certificate shows the result of checking configuration and operations within the stated scope against these criteria as of the assessment completion date. It does not guarantee that no incident will occur in the future. It is not a certification under ISO/IEC 27001, the PrivacyMark, or any other certification scheme.
Revision policy
The criteria are reviewed once a year, and changes are published on this page. Wording corrections and the addition of Recommended items (for example, v1.1) do not affect Certificates already issued. When Required items are added or changed (for example, v2.0), Certificates already issued remain valid under the previous version until they expire, and renewals are assessed against the new version.