資料8
参考資料1-2
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
Principle-Code for Protection of intellectual property and transparency for the
appropriate use of generative AI (provisional title) (draft)
1. Overview
(1) Basic Concept (Purpose)
This document is, based on the intent of the Act on Promotion of Research, Development
and Utilization of Artificial Intelligence-Related Technologies (Act No. 53 of 2025),
seeing the efforts of the EU AI Act (measures to ensure transparency and measures to
protect copyright) and the efforts in the field of corporate governance such as the
Stewardship Code (comply or explain), establishes the principles for measures to ensure
transparency and protect intellectual property rights that generative AI businesses
should take, thereby aiming to ensure a safe and secure usage environment for rights
holders and users, in order to balance the promotion of advances in generative AI
technology with the appropriate protection of intellectual property rights.
(2) Subjects to which this document applies
This document applies to "Generation AI Developers" and "Generation AI Providers"
(hereinafter collectively referred to as "Generation AI Businesses").
○
A "generative AI developer" is a person ( regardless of purpose or whether they are
a corporation or an individual ) who is responsible for building a generative AI
system (hereinafter collectively referred to as a "generative AI system") including
a generative AI model, the system infrastructure of the generative AI model, input
/output functions, etc. through the development of generative AI models and
algorithms, data collection (including purchase), preprocessing, generative AI model
learning and verification, and who has provided all or part of the generative AI
system related to the said development to the public (meaning an unspecified person
or a specified number of people; the same applies hereinafter).
○
A "generated AI provider" is a person (regardless of purpose or whether they are a
corporation or individual ) who is responsible for verifying generated AI systems,
implementing integration of generated AI systems with other systems, providing
generated AI systems or services, providing operational support for users of
generated AI systems to ensure normal operation, or operating generated AI services,
and who provides services to the public that incorporate generated AI systems into
applications, products, existing systems, business processes, etc. (hereinafter
collectively referred to as "generated AI services").
1
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
For clarity, the term "generative AI developer" in this document does not include a
person who uses data held by a single corporation or individual to provide a generative
AI system used exclusively by that person. Also, the term "generative AI provider" in
this document does not include a person who provides a generative AI service equipped
with a generative AI system specialized using data held by a single corporation or
individual exclusively to that person.
In addition , even if a generation AI business does not have its head office or main
office in Japan, this document shall apply if the generation AI system or generation AI
service is provided to Japan (including, but not limited to, when it is available to
Japanese nationals).
(3) Methodology adopted in this document
This document was established taking into consideration the circumstances and
respective intentions of generation AI businesses, generation AI users, and rights
holders, and does not require generation AI businesses to forcibly disclose information
belonging to them (including, but not limited to, trade secrets), but rather requires
them to comply with the principles set out below using a comply-or-explain approach.
The "comply or explain" approach requires businesses to either implement the principles
or, if they do not, explain the reasons for doing so. In other words, if a business
considers it inappropriate to implement any of the principles listed below in light of
its individual circumstances, it is possible to choose not to implement some of the
principles by providing a full explanation of the "reasons for not implementing" them.
Naturally, however, when providing such explanations, business operators should devise
ways to ensure that users and rights holders fully understand their response to the
principles they do not implement.
In addition, while implementing the principles, it is also considered beneficial to
actively explain the specific initiatives being undertaken by a company in order to gain
a sufficient understanding from users and rights holders.
2
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
(4) Visualization of the acceptance status of this document
To visualize the acceptance status of this document, the following is expected to the
Generative AI businesses that accept the principles set out below:
⚫
These businesses shall publish the following items on its own managed and operated
corporate website (a website that transmits official information such as an
overview of the AI generation business, business details, and product information)
or any other website with equivalent functions (hereinafter collectively referred
to as "corporate website, etc."), and submit the same in accordance with the
format prescribed by the Intellectual Property Strategy Promotion Office of the
Cabinet Office.
➢
Acceptance of the principles set out in this document (Acceptance Statement)
➢
The following matters regarding each principle set forth in this document
Implementation of each principle
If there are any principles that are not implemented, an explanation of
the reasons
⚫
Each item shall be reviewed and updated annually (and any updates shall be made
public).
The Cabinet Office Intellectual Property Strategy Headquarters will prepare a
reference form for comply or explain based on this document, publish a list of businesses
that have submitted notifications and links to the corporate websites and other
information published by those businesses, and encourage industries to proactively
submit notifications, with the cooperation of relevant ministries and organizations.
However, the Cabinet Office Intellectual Property Strategy Headquarters will not review
the content of the notifications, nor will it respond to inquiries from third parties.
3
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
2. Principles and exceptions set out in this document
(1) Principles set out in this document
[Principle 1]
The Generation AI Businesses shall disclose an outline of each of the matters set
out in (1) and (2) below (hereinafter collectively referred to as the "Outline
Disclosure Subject Matter") on the corporate website (a website that transmits
official information such as the outline of the Generation AI Businesses, business
details, and product information, which is accessible to all) that it manages and
operates, or on any other website with equivalent functionality, and make it accessible
to all, including users and rights holders.
(1) Transparency measures
The following items shall be disclosed:
A. Usage model
◼
Name (identifier, version, etc.)
◼
History including publication date (past versions, revision history,
etc.)
◼
Architecture and design specifications (status of licenses contracted
with third parties for model development, hardware, software and licenses
required for use, etc.)
◼
Terms of use (clarification of intended uses, restricted and prohibited
uses, etc.)
◼
Details of the model training process (training method, parameter settings
including inference process and decision basis, etc.)
B. Learning data
◼
Matters related to the data used for training and validation (type of
data, matters related to private datasets obtained by web crawling or
third parties , matters related to public datasets , matters related to
data collected by other means , whether synthetic data is used and for
what purpose, etc.)
◼
Crawler (purpose, data collection period, name/identifier, whether or not
a third-party crawler is used and its name/identifier, etc.)
C. Accountability
◼
The content of the state in which decisions made during the development,
provision, and use of generative AI systems or services can be traced and
traced to the extent technically possible and reasonable ( improving
traceability, clarifying who is responsible, allocating responsibilities
4
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
among
parties
involved,
responding
specifically
to
stakeholders,
documentation, etc.)
(2) Measures to protect intellectual property rights
The status of response to the following matters shall be disclosed.
◼
In order to ensure proper compliance with rights, principles for
protecting intellectual property rights will be established, and the
system of responsibility will be clarified. These will be reviewed at
least once a year, and a summary will be made public.
◼
When utilizing data, including for the development and training of
generative AI, these businesses shall not infringe on the intellectual
property rights of others.
◼
Respect access restrictions such as paywalls and employ crawlers that
follow machine -readable instructions such as robots.txt, ensure that
rights holders take appropriate measures, and the above measures shall be
published for each user agent and any changes must be notified.
◼
The learned logs are kept 1for a certain period of time .
◼
Work to avoid crawling so -called pirated sites .
◼
To the extent possible, technological measures shall be taken to prevent
the creation of products that infringe intellectual property rights .
◼
Wherever possible, use digital watermarks, C2PA and other technical
measures to verify the origin and provenance of content .
◼
Inform users that if they believe that a product infringes the
intellectual property rights of others, they should not use it.
◼
In order to ensure timely and appropriate relief for rights holders,
appropriate contact points will be established, including by utilizing
existing systems, application requirements will be made as clear as
possible, and records of responses will be kept.
(details)
1
Regarding the storage of logs, page 18 of the Ministry of Internal Affairs and Communications and
the Ministry of Economy, Trade and Industry's " AI Business Guidelines (Version 1.1)" states, under
"1. Ensuring verifiability, " that " In order to ensure the verifiability of AI decisions, logs of the
development process of AI systems and services, input and output during use , the AI learning process,
inference process , and the basis for decisions, etc., shall be recorded and stored within a reasonable
range in light of the amount and content of data." and " When recording and storing logs, the recording
method, frequency, storage period, etc. shall be considered in light of the characteristics and uses
of the technology used , taking into account the importance of investigating the cause of accidents,
etc. , considering measures to prevent recurrence, and proving the requirements for liability for
damages ."
5
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
○
It is expected that the accumulated efforts of each AI provider under Principle
1 will lead to the standardization of disclosed information, deepen mutual
understanding between AI providers, AI users, and rights holders, support
appropriate understanding and independent judgment among stakeholders, and
contribute to the creation of trustworthy AI. Therefore, when disclosing summary
information based on Principle 1, it is important to identify the presence and
scope of information that falls under each summary disclosure item while taking
into account the intent of "1. General Discussion," and to proceed with the
procedure while taking care not to violate other laws and regulations. Furthermore,
there is nothing to prevent AI providers from voluntarily disclosing details
regarding summary disclosure items.
○
For the level of detail that should be disclosed for each item subject to summary
disclosure, please refer to the specific examples of items subject to summary
disclosure published separately by the Cabinet Office Intellectual Property
Strategy Promotion Office.
6
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
[Principle 2]
In the event that a request for disclosure is made by a person who is currently taking
or preparing to take legal action, such as filing a lawsuit, filing a request for
mediation, ADR (alternative dispute resolution), or other legal procedure to realize
his or her rights or legally protected interests, or by an attorney authorized by
such a person or an agent authorized by law to take judicial action (hereinafter
referred to as the "Principle 2 Disclosure Requester"), regarding the following [Items
that may be Requested for Disclosure], and the request for disclosure satisfies all
of the items in the [Items that Disclosure Requests Must Satisfy] below, the Generation
AI Business Operator will respond to the request for disclosure.
[Items that may be requested for Disclosure]
◼
the data used for learning and validation (including, but not limited to, nonpublic datasets obtained by web crawling or third parties, public datasets,
datasets including synthetic data , and other data collected by other means )
includes information such as URLs that the AI generator queries (limited to
information that can be easily accessed and confirmed by the generator;
hereinafter referred to as "Principle 2 Control Information").
◼
If the party requesting disclosure is a generation AI provider and the
generation AI provider is unable to respond, the name of the person who
developed the generation AI model installed in the generation AI service
[Items that Disclosure Requests must satisfy]
①
Showing the person that the person is identified as a “Principle 2
Dicslosure Requester”
②
The purpose of use of the response to the request for disclosure is clearly
stated, and the person requesting disclosure pledges that they will not use
the response for any other purpose.
③
It indicates a specific Principle 2 Control Information, and specifies the
reasons for requesting disclosure from the AI generating company in relation
to the said Principle 2 Control Information.
7
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
(details)
○
Typical examples of disclosure requirements under Principle 2 are as follows:
◼
When a person who has created a work and posted it on website A discovers
an AI-generated product that is identical or similar to the work in
question, they provide the AI-generated business with the URL of the
page where the work is posted on website A and request disclosure of
whether the domain is included in the crawler crawl targets or is included
in the source of learning data provided by a third party.
○
With regard to the level of "reason that a person is identified as a Principle 2
Disclosure Requester" as set forth in ①, the generation AI business that receives
the request is required to show sufficient reason to believe that the person
falls under " a person who is currently taking or preparing to take legal action,
such as filing a lawsuit, filing a request for mediation, ADR (alternative dispute
resolution), or other legal procedure to realize his or her rights or legally
protected interests, or by an attorney authorized by such a person or an agent
authorized by law to take judicial action".
○
Principle 2: AI generators are expected to make efforts to provide as detailed
and easy-to-understand disclosure as possible so as not to impede the rights or
legally protected interests of the requester. Even in cases where the matters
requested for disclosure are considered to be trade secrets, they are expected
to first seriously consider and discuss the matter.
○
It is desirable for generation AI businesses to clarify and publicly announce
their own response policies for requests for disclosure from disclosure requesters
under Principle 2, taking into account technical challenges, costs, the purpose
of use, the content of legal procedures , etc., and making rational decisions to
avoid excessive burdens. (Note that generation AI businesses should consider
their own response policies in the event that they become the party to legal
procedures, as this may be related to the response strategy for those legal
procedures, etc.)
By accumulating such diverse comply-or-explain cases, it is expected that
businesses that are taking outstanding initiatives will be appropriately
evaluated based on market principles, and that mutual understanding will deepen
between AI generation businesses, AI generation users, and rights holders.
8
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
○
When implementing Principle 2, measures may be taken to prevent abusive requests,
such as imposing a certain fee or a limit on the number of requests within a
reasonable period of time . However, care must be taken to avoid imposing fees
or limiting the number of requests that discourage, make difficult, or discourage
disclosure requests.
○
Although no specific rules are set forth regarding the timing of responses,
Principle 2 states that efforts should be made to disclose information promptly
within a reasonable period of time so as not to impede the realization of the
rights or legally protected interests of the person making the disclosure request.
○
From the perspective of fulfilling accountability as a provider of generative AI
systems or generative AI services to the public, simply stating that a system for
implementing Principle 2 has not been established is not sufficient as an
"explanation," and the company shall provide an appropriate explanation as to
when the establishment of such a system will be completed, taking into account
factors such as the scale of its business as a business operator.
○ Other methods of collecting information after the filing of a lawsuit, other than
requests for disclosure based on Principle 2, include inquiries between the
parties (Article 163 of the Code of Civil Procedure) and applications for document
production orders (Article 221 of the Code of Civil Procedure).
9
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
[Principle 3]
If a person (hereinafter referred to as the “Principle 3 Disclosure Requester”)who
uses a Generative AI system or Generative AI service provided by a Generative AI
Business Operator to generate movies, music, plays, literature, photographs, manga,
animation, computer games, or other text, figures, colors, sounds, actions, or images,
or combinations of these, or programs for providing information related to these via
a computer (meaning instructions to a computer combined to produce a single result)
requests disclosure of the items listed in [Items that may be requested for disclosure]
below, and requests for disclosure satisfies all of the items in [Items that disclosure
requests must satisfy] below, the Generative AI Business Operator will respond to the
request for disclosure.
[Items that may be requested for disclosure]
◼
Whether the data used for learning and validation (which refers to, but is not
limited to, non-public datasets obtained by web crawling or third parties,
public datasets, datasets including synthetic data, or data collected by other
means ) contains information such as URLs (limited to those that can be easily
accessed and confirmed by the generation AI business operator; hereinafter
referred to as "Principle 3 Control Information") that contain content (which
refers to "content" as defined in Article 2, Paragraph 1 of the Act on Promotion
of
Creation,
Protection
and Exploitation of
Content; the
same applies
hereinafter) that is identical or similar to the product generated by the
Principle 3 Disclosure Requester.
◼
If the party requesting disclosure is a generation AI provider and the generation
AI provider is unable to respond, the name of the person who developed the
generation AI model installed in the generation AI service
[Items that Disclosure Requests must satisfy]
① The reason for identifying as a Principle 3 Disclosure Requester includes showing
that the product generated by the person requesting disclosure and the prompt
used to generate the product are provided.
② The purpose of use of the response to the request for disclosure is clearly
stated, and the person requesting disclosure pledges that the response will
not be used for any other purpose than the stated purpose or for the purpose
of filing a lawsuit, filing for arbitration, ADR (alternative dispute
resolution), or other legal procedures.
10
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
③ It indicates a specific Principle 3 Control Information, and specifies the
reasons for requesting disclosure from the AI generation business operator in
relation to the said Principle 3 Control Information.
(details)
○
Typical examples of disclosure requirements under Principle 3 are as follows:
◼
When a person who has created a generated AI product using generation AI
service A, which is capable of generating images, discovers that an image
identical or similar to the generated AI product is posted on website B,
the person requests the generation AI provider of generation AI service
A to disclose the generated AI product, the prompt used to generate the
generated AI product, the purpose of use of the generated AI product,
and the URL of website B, and to ask whether the domain part of the URL
is included in the crawl targets for learning data when developing the
generation AI system installed in generation AI service A, or whether it
is included in the source of learning data provided by a third party, or
if the generation AI provider is unable to answer, to disclose the name
of the person who developed the generation AI model installed in the
generation AI service.
○
In order to avoid hindering the utilization of generated AI products by those
requesting disclosure, the generating AI business operator is required to make
efforts to provide as detailed and easy-to-understand disclosure as possible.
Furthermore, even in cases where the matter related to the request for disclosure
is considered to be a trade secret, it is expected that the business operator
will first seriously consider and discuss the matter.
○
It is desirable for generation AI businesses to clarify and publicly announce
their own response policies for requests for disclosure from disclosure requesters
under Principle 3, taking into account technical challenges, costs, the purpose
of use, the content of legal procedures, etc., and making rational decisions to
avoid excessive burdens. (Note that generation AI businesses should consider
their response in the event that they become the party to legal procedures, as
this may be related to the response strategy for those legal procedures, etc.)
By accumulating such diverse comply-or-explain cases, it is expected that
businesses that are taking outstanding initiatives will be appropriately
11
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
evaluated based on market principles, and that mutual understanding will deepen
between AI generation businesses, AI generation users, and rights holders.
○
When implementing Principle 3, measures may be taken to prevent abusive requests,
such as imposing a certain fee or a limit on the number of requests within a
reasonable period of time. However, care must be taken to avoid imposing fees or
limiting the number of requests that discourage, make difficult, or discourage
disclosure requests.
○
Although no specific rules are set out regarding the timing of disclosure,
companies are expected to make efforts to make prompt disclosure within a
reasonable period of time.
○
From the perspective of fulfilling accountability as a provider of AI generating
systems or AI generating services to the public, simply stating that a system for
implementing these principles has not been established is not sufficient as an
"explanation," and the company will provide an appropriate explanation as to when
the establishment of such a system will be completed, taking into consideration
factors such as the scale of its business as a business operator.
○ Other methods of collecting information after the filing of a lawsuit, other than
requests for disclosure based on Principle 3, include inquiries between the
parties (Article 163 of the Code of Civil Procedure) and applications for document
production orders (Article 221 of the Code of Civil Procedure).
(2) Exceptions to the principles outlined in this document
Some Generative AI Providers develop Generative AI systems or provide Generative AI
services using open source software. As a result, it is anticipated that there may be
cases where it is difficult to disclose or explain some of the summary disclosure items
under Principle 1, or to disclose items that can be required to be disclosed under
Principles 2 and 3. Therefore, this document provides the following exceptions.
[Exceptions to Principles 1 to 3]
Among generation AI businesses engaged in the development and learning phase (preand post-learning), those that use open source software to conduct all or part of
their business and for which it is difficult to both disclose and explain some of the
matters subject to summary disclosure under Principle 1 due to their use of open
source software, and those for which it is difficult to disclose matters that may be
required to be disclosed under Principles 2 and 3, may instead disclose those matters
12
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
by clarifying the fact that they use open source software and the details of the
license for that open source software.
(3) Points to note when selecting "Explain"
Even if a generative AI business does not implement all or part of the principles and
"explains" the reasons for not implementing them, in light of the purpose of this
document, which is to ensure a safe and secure usage environment for users in order to
promote the advancement of generative AI technology while also properly protecting
intellectual property rights, the businesses should devise ways to ensure that users
and rights holders fully understand its response to the principles it does not implement.
Furthermore, even if AI businesses have declared that they will implement each
principle set forth in this document (i.e., has declared its acceptance), if it is
determined that they are not actually implementing each principle set forth in this
document, such as by narrowing the scope of disclosure based on provisions in terms of
use, etc., a separate "explain" is required. For clarification , it is not sufficient
to simply show that there is a provision in the contract or terms of use that abolishes
or restricts the application of the principles set forth in this document (an override
clause); it is necessary to explain why that provision has been included to abolish or
restrict the application.
(4) Other matters
The government is expected to evaluate the disclosures and specific initiatives of AI
businesses and provide certain incentives in the various projects and systems that it
implements and operates.
Furthermore, this document will be revised based on the results of consideration of
the response status of AI generation businesses and trends in international efforts,
etc., if deemed necessary.
End
13
Note: This is a provisional translation, and the Cabinet Office is not responsible for its content.
For official information, please refer to the Japanese version.
[Notes]
The following items will be further considered:
○
The items and details of the "Outline Disclosure Subject Matter" set forth in
Principle 1 and the "Items that May be Required for Disclosure" set forth in
Principles 2 and 3
○
Granularity of disclosure related to each principle
○
The status of Principles 2 and 3
○
Considerations for startups, seeing that this document is aimed at those who
provide generative AI systems and generative AI services to the public
○
Measures for disseminating and incentivizing each principle among AI generation
businesses that are subject to this document (including AI generation businesses
that do not have their head office or main office in Japan)
14
資料10
参考資料2-2
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
Principles-Code for Protection of intellectual property and transparency for
the appropriate use of generative AI (provisional title) (draft)
Summary of disclosure items Specific examples
*These specific examples are provided by the Secretariat as examples of
disclosures made by businesses under the Code for reference purposes, so
that businesses can refer to them when submitting their opinions in the
public comment period.
*As an example of what would happen if you selected "Explain" for a matter
to be disclosed, the secretariat has provided "Examples of when to
explain" for some matters for reference.
classification
item
Name (identifier,
version, etc.)
Specific examples
○○ Ver2.1
History including
publication date (past
Released on 〇/〇/〇, fixed △△ function on
versions, revision
△/△/△.
history, etc.)
[Architecture/design specifications]
Transformer architecture
[Status of license contract with third party
Usage
Architecture and
model
design specifications
related
(status of licenses
contracted with third
parties for model
development,
hardware, software
and licenses required
for use, etc.)
for model development]
Contract signed with XX company
<Example of when to explain>
In developing the model, we have concluded
a license agreement for the model used with
a third party, but due to confidentiality
obligations under the agreement, and
expressing that it is at the core of our
businesses, we cannot disclose the name of
the third party.
[Hardware and software required for use]
GPU with a capacity of ○○ GB or more, △△
Ver. 3.1 or higher
1
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
[License policy for users]
After publishing and sharing the model , we
ask that you agree to a free open source
license that allows the provider to freely
access, use, modify, and redistribute the
model or its modified versions under certain
conditions.
*License policy (URL of reference page)
Terms of use
(clarification of
intended uses,
restricted and
Prohibited uses: Violent, sexual, or CBRNrelated uses, or any use that violates any law
or regulation
* URL of the reference page
prohibited uses, etc.)
The pre-training stage is optimized based on
Details of the model
training process
(training method,
parameter settings
including inference
process and decision
basis, etc.)
gradient descent. The model employs an
autoregressive language model and is trained
to predict the next word using a large corpus
of trillions of tokens. Post-training is then
performed using techniques such as
reinforcement learning (RLHF) using a
dataset that reflects human preferences,
adjusting the response to match human
values and intentions.
2
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
classification
item
Specific examples
Matters related to the
Learning
data
data used for training
[Data type]
and validation (type
Text, images, audio, video
of data, matters
[Web crawl]
related to private
Conducted
datasets obtained by
[Matters related to private datasets]
web crawling or third
Obtained through license agreement
parties, matters
[Matters related to public datasets]
related to public
Uses ○○ structured dataset
datasets, matters
[Matters related to data collected by
related to data
other means]
collected by other
Obtained from data provider △△
means, whether
[Whether synthetic data is
synthetic data is used
used/purpose]
and for what purpose,
Uses synthetic data/enhanced security
etc.)
[Purpose]
Collect data to improve the model
[Name/Identifier: Data collection period]
Crawler (purpose,
○○bot: Collected continuously from 〇/〇/
data collection period,
〇
name/identifier,
△△bot: Collected continuously from △/△/
whether or not a
△
third-party crawler is
[Third-party crawler name/Identifier:
used and its
Data collection period]
name/identifier, etc.)
●●bot: Collected continuously from ●/●/
●
▲▲bot: Collected continuously from ▲/▲/
▲
3
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
classification
item
Specific examples
[Traceability]
Ensuring that
Model evaluation logs will be continuously
decisions made during recorded and retained for the period
the development,
specified by the rules of ○○ Company.
provision, and use of
[Clearance of responsible person]
generative AI systems The CAIO of ○○ company will be the
or services can be
person in charge and will be responsible for
tracked and traced to
the work of the ○○ team at the company.
the extent technically
[Distribution of responsibility among
possible and
parties involved]
Accountability reasonable
In accordance with the above contract, we
(improving
will clarify the allocation of responsibilities
traceability, clarifying
between the rights holder and users, etc.,
who is responsible,
and confirm the agreement with the other
allocating
party.
Relationships
responsibilities among [Specific responses to stakeholders]
parties involved,
According to ○○ company's ○○
responding specifically regulations
to stakeholders, and
[Documentation]
documenting such
Documented according to ○○ company's
information)
○○ regulations
4
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
classification
item
Specific examples
In order to ensure
proper compliance
with rights, principles
for protecting
・○○ principles for intellectual property
protection have been established.
intellectual property
*Outline of ○○ principles (URL of
rights will be
established and the
system of
responsibility will be
Measures to
clarified. These will be
protect
reviewed at least once
intellectual
a year, and a
property
summary will be
rights
reference page)
・Establish a ○○ department to comply with
the ○○ principles, review the results once a
year, and publish the results on the
company's website (URL of the reference
page)
made public.
When utilizing data,
・Establish a ○○ process to ensure that
including for the
development and learning do not infringe on
development and
the intellectual property of other companies
training of generative
*Outline of the ○○ process (URL of
AI, these businesses
reference page)
shall not infringe on
・Regarding use, include clauses regarding the
the intellectual
prevention of infringement of rights in the
property rights of
terms of use
others.
* URL of reference page
5
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
classification
item
Specific examples
to respect access
restrictions such as
paywalls and employ
crawlers that follow
machine-readable
instructions such as
Comply with paywalls and use crawlers that
robots.txt . To ensure
comply with robots.txt.
that rights holders
Publish the policy, including the above, on
take appropriate
your company website (URL of the reference
measures, the above
page).
measures must be
published with user
Measures to
protect
intellectual
agents and any
changes must be
notified.
property
rights
The learned logs are
kept for a certain
Keep logs for the duration of use
period of time.
・In order to exclude pirated sites from the
collection of learning data, when information
Work to avoid
crawling so-called
pirated sites.
about pirated sites is provided by public
institutions, etc., we will review the content
and take measures such as excluding them.
・Exchange information with rights holder
organizations and make efforts to identify
pirated sites
6
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
classification
item
Specific examples
To the extent
possible, technological
measures shall be
taken to prevent the
creation of products
that infringe
Equipped with a filtering function to prevent
output of products that infringe intellectual
property rights
intellectual property
rights.
Wherever possible,
use digital
watermarks, C2PA
and other technical
measures to verify the
C2PA compliant , digital watermark
technology implemented
origin and provenance
of content.
Measures to
Inform users that if
protect
they believe that a
intellectual
product infringes the
Include clauses in the terms of use to prevent
property
intellectual property
infringement of intellectual property rights
rights
rights of others, they
should not use it.
In order to ensure
timely and
appropriate relief for
rights holders,
appropriate contact
points will be
established, including
by utilizing existing
systems, application
Establishment of copyright contact point (URL
of reference page), keeping correspondence
records
requirements will be
made as clear as
possible, and records
of responses will be
kept.
7
This is a provisional translation, and the Cabinet Office is not responsible for its content. For
official information, please refer to the Japanese version.
8