AI Policy - EXTERNAL

Development
Chapter
SAP
AI Ethics
Handbook
Handbook for applying
SAP’s Global AI Ethics Policy
across the AI Factory Process
External
Content.
SAP Definition of AI 0.6
Ethical AI Guiding Principles 0.8
The AI Ethics Policy 0.10
Red Lines 0.12
AI Factory Process 0.18
A Human Centered Approach 0.20
01. Ideation 0.24

SAP.com
02. Validation 0.28
Authors
Camila Lombana Diaz,
Saskia Welsch,
03. Realization 0.34
Sebastian Wieczorek
04. Productization 0.38
Special Thanks to:
Claus Holzknecht, Detlef Plümper, 05. Operations 0.42
Jean-Christophe Pazzaglia,
Julia Schöndienst, Nicole Sanchez Cordero
Pavithra GK, Philip John, Rainer Kerth,
Vikram Nagendra.
© 2023 SAP SE or an SAP affiliate company.
Updated version: August 2023

All rights reserved
https://www.sap.com/about/legal/copyright.html
2 SAP AI Ethics Handbook | External

Foreword.
We at SAP believe that Artificial The SAP Global AI Ethics Policy
Intelligence (AI) has great potential to ensures the development and
create opportunities for businesses, deployment of SAP’s AI systems in line
governments, and societies. But for with the established guiding principles
people to trust in AI solutions the and core organizational values. This
development, deployment, use, and AI Ethics Handbook will guide SAP
sale of AI systems must be governed employees in implementing the AI
Disclaimer by clear moral rules. Ethics Policy during all stages of the
AI development lifecycle. All policy
In addition to this handbook, To ensure adherence and account- requirements are mapped to the
employees should regularly check ability, SAP has made ‘Trustworthy respective stages of SAP’s AI Factory
on and must adhere to other existing AI’ a key pillar of its strategy and processes.
policies such as Data Protection has committed to guaranteeing the
and Privacy (DPP) policy as well as sustainable and safe development My sincere gratitude goes to
applicable product development of all AI systems. This promise SAP all people who have helped shape
standards. keeps by adhering to a governance Trustworthy AI at SAP and set up
structure, which ensures the ethical the structures and instruments for
This document is intended to serve as and transparent development of AI implementing it. This handbook
a general guide for applying the SAP solutions. contributes to these efforts and
Global AI Ethics Policy in everyday work enables and supports employees in
and is not meant to advise, implement, The foundation of SAP’s approach to developing ethical, trustworthy AI at
or replace either existing or new AI Ethics is our pledge to uphold and SAP.
processes, policies, or procedures. support the Universal Declaration
This guide is not intended to be of Human Rights. An essential part I would like to encourage each
comprehensive and is not a substitute of this is prohibiting discrimination development colleague to get familiar
for legal or data protection and privacy and harassment of humans based with the handbook and address any
advice. on personal factors, such as race, questions you might have related to
ethnicity, religion, age, and gender. the processes to ai.ethics@sap.com.
To the extent that a 3rd Party AI system ‘Trustworthy AI’ at SAP is a pledge to Together we can ensure compliance
(e.g., ‘TensorFlow’) is embedded in SAP safeguard our customers and users with ethical development standards
solutions, the requirements laid down from harm, treating all individuals fair and establish SAP’s position as a
in this handbook shall apply to the and just. thought leader in this field.
overall SAP software solution.
Dr. Feiyu Xu
Global Head of Artificial Intelligence, SAP
Senior Vice President
4 SAP AI Ethics Handbook | External

SAP differentiates between two types
SAP’s
of AI systems:
Definition
Rule-Based AI
• Rule-based AI systems are characterized by the fact that
of AI
their behavior is fully defined by rules created by human
experts. These systems are often described as symbolic
or expert systems.
Learning-Based AI
AI is typically defined as the ability • Learning-based AI systems are differentiating themselves
of a machine to perform cognitive by the fact that humans define the problem and the goal,
functions we associate with human but the behavior, rules and relationships required for the
minds, such as perceiving, reasoning, system are learned in an automatize way. With the help of
learning, and problem-solving. data, they train how to solve a problem and continuously
adapt their function in this process.
It requires a system to correctly
interpret external data, learn from At SAP, Risk Assessments shall be performed on such
such data, and use those learnings to Learning-based AI systems during the development phase
achieve specific goals through flexible and subsequent phases to ensure that there are
adaptation. no unintended biases.
6 SAP AI Ethics Handbook | External Development Chapter 7

SAP Ethical AI
uiding
G General
Principles. Guidelines.
01 . How this handbook is
We are driven by our values structured
SAP’s AI Ethics Guiding Principles, and SAP personnel shall approach ethical
02 . AI Ethics Policy are presented in the dilemmas and trade-offs related
to their use via reasoned, context-
We design for people first two chapters.
relevant, and evidence-based decision
The “Red Lines” chapter clarifies which making rather than intuition or random
AI use-cases are prohibited within SAP. discretion.
03 .
It includes the definition of high-risk
We enable businesses beyond bias SAP employees should raise any issues
use-cases pointing out the necessary
processes for approval. related to possible policy breaches for
evaluation by their immediate L1 unit,
04 . SAP’s AI Factory Process chapter even if employees only have doubts or
We strive for transparency and integrity allows for an overview of its different concerns.
stages, including the people involved
in the process. Before each stage is If questions or concerns remain, or a
05 . introduced in more detail on how the decision is unable to be made by the
AI Ethics Policy is applicable, some appropriate L1 unit, then employees
We uphold quality and safety standards
general guidelines are described. At should inform the AI Ethics Office
the end of the handbook, you can find via an e-mail to ai.ethics@sap.com
helpful links and sources. describing the use case.
06 .
We place data protection and privacy at our core
07 .
We engage with the wider societal challenges of AI

The AI Product
Ethics Policy. Standards
The AI Ethics Policy clarifies how SAP’s Guiding Principles The pillars of the policy have been linked to the existing
for Artificial Intelligence relate to AI use cases. The policy product standards to provide a transparent guidance.
applies to SAP and all its employees worldwide. It defines Please beware that any use case falling into the “Red-Line”
intent, expectations and obligations for employees involved category should be immediately dismissed.
in development, deployment and sale of AI systems.

01. 02. 03.

The Pillars of the Policy. Human
Addressing Transparency
Agency
Bias & & Explainability
&
Discrimination
Oversight
01. RED LINE

Human Agency Deanony-
& Oversight mization
DPP-350 (Product standard)
Human Intervention
Allow human intervention for all automated decision
02. processes. DPP-361
03.
Addressing Bias (Product
Transparency Transparency
standard)
& Discrimination & Explainability Provide transparency when using AI during processing
of personal data.
Apply adequate
Bias anonymization
Ensure used AI methods do not result in inaccuracies techniques
detrimental to data subjects’ rights.
Safeguards
Implement safeguards to protect personal data used
to train AI models.

Red Lines
01.
Personal
SAP Red Line Cases Freedom
Human Surveillance 02.

No SAP AI use cases should fall under the so-called Human surveillance that Society
Red Lines defined in the SAP AI Ethics Policy. These are is utilized for targeting 03.
defined as highly unethical under the purposes for AI. individuals or groups, Environment
either by biometrics, Manipulation
If your use case is built for these purposes, you are re- facial recognition, or other for deception or unfair
quired to immediately stop developing, deploying, and identifiable features, with manipulation of Environmental Harm
selling that use case. the purpose of disregard- individuals or groups AI system development
ing or abusing the rights of via public forums, or deployment shall be
SAP’s Red Lines are outlined on the right. the individuals or groups. media, or moderation conducted with
of other similar uses. minimum to no explicit
Discrimination damage to the
Purposes which cause Undermine Debate environment.
individuals or groups to systems which under-
be discriminated against mine human debate or
or excluded from equal democratic electoral
access to AI’s benefits and systems.
opportunities to the wider
population. Intentionally harmful
impacts on users and/
Deanonymization or those directly and
Deanonymization of indirectly affected by
already anonymized data the system.
which may result in the
identification of individuals
or groups.

High Risk
SAP High Risk Cases Determining High Risk Cases
SAP classifies AI use cases as high-risk under certain Processing personal data:
circumstances. We base these criteria on what kind of AI Does the use case process any information relating to an identified
has led to negative consequences for individuals or whole or identifiable natural person for training purposes or during productive
populations in the past (see, for example, Amazon’s case usage? – Use Cases with anonymized data sets or only the process of
of a discriminating AI recruiting tool). anonymizing Personal data does not qualify as High Risk Case.
High-risk use cases are not prohibited within SAP; Processing sensitive personal data:
however, they must first go through an assessment Is the use case including the processing of sensitive personal data
process of the AI Ethics Steering Committee before they like information on sexual orientation, religion, biometric data
can be further developed, deployed, and sold. (including face imaging &/or voice recognition)?
Please be mindful that further changes could be Automated decision making:

necessary after the assessment process of the Is the use case exhibiting fully or partially automated decision making?
AI Ethics Steering Committee. Does it include cases where no human intervention or human
supervision takes place? – This excludes any recommender system
Negatively affected individuals:

Could the use case negatively affect the social well-being of individuals
(health and safety), or intrude/restrict an individual’s fundamental
rights and freedom?
High-risk application:
Does the use case belong to one of the following domains:
E.g. Categorisation of natural persons, Management and operation of
critical infrastructure, Employment/HR, Healthcare, Private services and
Public services and benefits, Law Enforcement, Migration, Democratic
processes?

start
Risk Classification & Assessment Process
yes yes
Are you working Complete the
on a AI use case? Risk Classification form stop
yes 01. 02. 03.

Is it classified as a red Personal Society Environment
line case? Freedom
no
Red Line Case
yes
Is it a Generative AI
usecase?
Assessment
Steering Committee
no
yes
Is it processing sensitive Complete the AI Ethics
personal data? yes
Policy Self- Assessment
no
yes Automated decision
making? High Risk Case
Is it processing yes
yes
personal data? no
Does it negatively no
no affect individuals? High-risk sector?
no
Complete the AI Ethics

Policy Self- Assessment
Continue to
AI Factory
Process
Standard Case

The AI Factory
Process:
Use case
identification
01. based on common
Ideation domain and AI
expertise
SAP 02.
AI Factory
Validation
03.
Realization
Process
Continuous
Improvement
Experiments
to assess feasibility
Development
of AI functions
04.
Producti-
zation
05.
Operations
Integrating AI functions
into b
usiness process
Deliver embedded
AI functions
to customers

People Involved
A In SAP we believe that a greater emphasis and under-

standing on “Human-Centered AI” will strengthen the
Human
benefits of users and society in business, by reducing the
fears and/or existential threats our users may encounter.
Centered
Thus, our AI processes must be conducted through the
understanding of our user and stakeholder needs.
Approach
Methods of user research, design thinking, and correctly
defining user stories are mandatory aspects of our
development.
This enables that our products are as close as possible

to the real needs of our target groups avoiding
unconscious bias, or possible unintended harms.

AI Use Case Owner Designer
Responsible for the coordination of the delivery of an AI Provides the Human-Centered understanding of the
enabled product or feature end to end across all phases product by developing a user need strategy including
of the AI factory process. stakeholder, end users, and product roadmap.
Conducts: user research, flows, usability evaluation,
Factory Process Activity: including voice response and conversational user
interaction.
Factory Process Activity:
AI Data Scientist AI OPS Engineer

Develops, implements or applies AI methods to derive
Operates the AI enabled products or individual
solutions to business problems that can be translated
AI functions and manages life-cycle aspects like version
into AI functions.
updates.
Factory Process Activity: Factory Process Activity:
User Assistance Developer

AI Engineer
Responsible for designing, developing, reviewing, and
Responsible for the design, implementation and
maintaining content for product documentation, user
maintenance of AI functions of a product.
interface messages, and conversational applications.
Works according to specifications and project plans. Collaborates with internal stakeholders to design
customer-facing content and implements embedded help
delivery mechanisms that improve how technical informa-
tion is delivered to customers.
Product Developer Customer

Develops the business logic of a product or at least parts Represents the user of the AI enabled product.
of it. This task includes the integration of the provided AI As user he must be active in the ideation, validation, and
functions into the product context. operations phases.
Factory Process Activity: Factory Process Activity:

01
Ideation
AI Use Case AI Data Designer Customer We will follow
Owner Scientist
& document
these requirements:
Example:
A product owner within SuccessFactors has a deep
A.
understanding of his customers’ pain points. To help
them run efficiently, he identifies a business process that
can be automated. Before the team builds a prototype Human Agency & Oversight.
of the AI use case, they together collect all the relevant
information about the use case and discuss whether any
• Before implementation, the degree of • An appropriate governance mechanism
of SAP’s red lines are touched (e.g., AI used for mass freedom of the AI system must be defined. has to be chosen
surveillance) and identify risks of unethical behaviour
The decision-making degrees of freedom of the AI Human oversight shall be achieved through an
(E.g., automated decision-making). system must be defined. appropriate governance mechanism. This could include
but not be exclusive to human-in-the-loop, human-on-
• The target definition of the AI system the-loop, or human-in-command, and shall be decided
They fill out the use case risk assessment template and on a case-by-case basis.
must be given by a human.
send it to the AI Ethics Office for due diligence.
• Decisions by an AI system may always be • Human oversight must be introduced
overruled by a human. where humans are directly impacted.
AI systems shall be subject to appropriate human In situations where humans may be directly impacted
oversight, and the rights and freedoms of a human by a decision made by SAP’s AI system, human
shall exceed that of AI systems. oversight shall be introduced to safeguard that AI
system does not undermine human autonomy or
introduce unintended consequences.

01
Ideation
B. Addressing Bias & Discrimination
• Affected users should be involved in the

development process
Wherever possible, developers shall seek to involve
impacted/affected users to evaluate and check that
outputs are diverse and discrimination free.
• The AI system should address the widest

Transparency & Explainability
possible range of end-users.
AI software shall be user-centric, addressing the widest • AI systems interacting directly with
possible range of applicable end-users, and following humans must be identifiable as such.
relevant accessibility standards, regardless of users’
age, gender, abilities, or characteristics. Where applicable, when interacting directly with
humans (including via Conversational AI or ‘Chat-
bots’): AI systems shall be made identifiable as such to
appropriate end users. AI systems shall be developed
such that it does not encourage humans to develop
attachment and/or empathy of users towards the AI
system. AI systems shall clearly signal to end users
that its social interaction is simulated.
• Prevent misuse of the AI system in

production.
AI system development shall take into account the
context and environment in which the system will
operate such that, even with good intentions, no harm
or misuse is likely to occur to humans when AI systems
are deployed.

02
Validation
AI Use Case AI Data Designer Customer We will follow
Owner Scientist
& document
these requirements:
Example:
The SuccessFactors team moves forward with the
A.
use case considering the instructions of the AI Ethics
Steering. They are interacting with potential customers Human Agency & Oversight
and user groups to discuss expectations in regards to
the functional scope as well ethical considerations like
• Before implementation, the degree of • Human oversight must be introduced
transparency and human agency.
freedom of the AI system must be defined. where humans are directly impacted.
Before implementation, the decision-making degrees In situations where humans may be directly impacted
Based on their findings they carry out experiments and of freedom of the AI system must be defined. by a decision made by SAP’s AI system, human
build prototypes, iteratively aligning the results with their oversight shall be introduced to safeguard that AI
system does not undermine human autonomy or
stakeholders. The team makes sure that the data used introduce unintended consequences.
• Decisions by an AI system may always be
for exploration and experimentation is balanced and overruled by a human
representative and that the used AI methods are suitable
AI systems shall be subject to appropriate human • Clear and simple explanations have to be
for the desired level of explainability. oversight, and the rights and freedoms of a human
provided for automated decisions.
shall exceed that of AI systems.
As far as is practical, a clear and simple explanation
shall be provided as to how decisions were made by an
• An appropriate governance mechanism AI system used in automated decision processes.
has to be chosen.
Human oversight shall be achieved through an
appropriate governance mechanism. This could include
but not be exclusive to human-in-the-loop, human-on-
the-loop, or human-in-command, and shall be decided
on a case-by-case basis.

02
Validation
• Unfair bias must be avoided.

SAP shall endeavour to achieve fairness; AI systems
shall not intentionally generate unfairly biased outputs.
• Inclusive data must be used for training.

Where relevant, the data used to train AI systems shall
be as inclusive as possible, representing as diverse a • Model architectures must not include • The data sets and the development
cross-section of the population or past situations as ethically objectionable features or processes must be documented.
possible, and as free as possible from (or accounted
structures. The data sets and the processes that produce an AI
and mitigated for) any historic or socially constructed
system’s decisions, including those of data gathering
biases, inaccuracies, errors, and mistakes. The model architectures shall not include target
and data labelling as well as the algorithms used by the
variables, features, processes, or analytical structures
developed AI system, shall be documented to allow for
which are unreasonable, ethically objectionable, or
• Measures to detect bias must be realized. traceability and transparency.
unable to be validated according to the principles laid
out in this document.
SAP shall endeavour to detect unfairly biased outputs
and shall implement technical and/or organizational • AI systems interacting directly with
measures to prevent direct or indirect prejudice,
• The AI system should address the widest humans must be identifiable as such.
discrimination, or marginalization of groups or
individuals, e.g. by reducing bias in training data. possible range of end-users.
Where applicable, when interacting directly with
AI software shall be user-centric, addressing the widest humans (including via Conversational AI or ‘Chat-
possible range of applicable end-users, and following bots’): AI systems shall be made identifiable as such to
• Affected users should be involved in the relevant accessibility standards, regardless of users’ appropriate end users. AI systems shall be developed
development process. age, gender, abilities, or characteristics. such that it does not encourage humans to develop
Wherever possible, developers shall seek to involve
system. AI systems shall clearly signal to end users that
impacted/affected users to evaluate and check that
• The use of data for testing must comply its social interaction is simulated.
outputs are diverse and discrimination free.
with DPP Policy.
The use of data for the testing of AI systems shall
• Data used for training and testing has to comply with applicable data protection, and privacy • Prevent misuse of the AI system in
be representative and generalizable. laws. production.
It shall be trained and tested on as expansive as AI system development shall take into account the
is feasible, representative, relevant, accurate, and context and environment in which the system will
generalizable datasets. operate such that, even with good intentions, no harm
or misuse is likely to occur to humans when AI systems
are deployed.

02
Validation
C. Transparency and Explainability
• AI systems interacting directly with

humans must be identifiable as such.
Where applicable, when interacting directly with
humans (including via Conversational AI or ‘Chat-
bots’): AI systems shall be made identifiable as such to
appropriate end users. AI systems shall be developed
such that it does not encourage humans to develop
system. AI systems shall clearly signal to end users
that its social interaction is simulated.

03
Realization
AI Use Case AI Data AI
Owner Scientist Engineer
We will follow
& document
these requirements:
Designer User Assistant
A.
Developer
Human Agency & Oversight

Example:
• Additional testing is necessary in case of • AI systems automating decisions must
The SuccessFactors team has successfully built a composed AI systems. be tested extensively to avoid unintended
prototype and now enters the realization stage. After When two or more AI systems are connected to each behaviour.
the model has been trained, the data scientists and other or embedded within each other, additional testing When human-on-the-loop models are used,
engineers explore a technical solution to provide local and control measures should be performed for the
individual AI systems as well as for the overall system.
appropriate extensive testing and governance shall
be conducted during development and deployment
explainability. This way, they can provide the factors to ensure the system behaves as intended by the
behind a particular decision by the AI system for the user. • Clear and simple explanations have to be developers and does not have any unintended
provided for automated decisions. behaviour, outputs, or usage.
As far as is practical, a clear and simple explanation

AI system used in automated decision processes.

03
Realization

• Model architectures must not include
ethically objectionable features or
shall not intentionally generate unfairly biased outputs. structures.
C.
The model architectures shall not include target
• Inclusive data must be used for training. variables, features, processes, or analytical structures
be as inclusive as possible, representing as diverse a
cross-section of the population or past situations as
and mitigated for) any historic or socially constructed • The data sets and the development
biases, inaccuracies, errors, and mistakes.
• For decisions about affecting humans
• Where feasible, a fairness function shall processes must be documented. explanations have to be provided to the data
be used. subject.
The data sets and the processes that produce an AI
• Measures to detect bias must be realized. Where feasible, a fairness function shall be applied to system’s decisions, including those of data gathering In alignment and compliance with applicable data
test AI systems for unbiased output. and data labelling as well as the algorithms used by the
SAP shall endeavour to detect unfairly biased outputs protection and privacy law, AI systems that engage in
and shall implement technical and/or organizational developed AI system, shall be documented to allow for profiling or automated decision-making must be able
measures to prevent direct or indirect prejudice, traceability and transparency. to provide explanations to the extent possible to data
discrimination, or marginalization of groups or • The AI system should addresses the subjects upon request, describing the data segment
individuals, e.g. by reducing bias in training data. widest possible range of end-users. the subject was placed into and the reasons they were
AI software shall be user-centric, addressing the widest • The AI system’s capabilities and limita- placed there.
possible range of applicable end-users, and following tions must be documented. In addition, the reasons as to why the decision was
• Affected users should be involved in the relevant accessibility standards, regardless of users’
The capabilities and limitations shall be documented made shall be provided if requested by the data
development process. age, gender, abilities, or characteristics. subject. The explanation must be such as to provide
as part of the development process in a manner
Wherever possible, developers shall seek to involve appropriate to the use case at hand. This shall include the data subject grounds to challenge the decision.
impacted/affected users to evaluate and check that information regarding the AI system’s level of accuracy
outputs are diverse and discrimination free. • The use of data for testing must comply (performance metric), as well as its limitations and
with DPP Policy. capabilities. • The methods used for development,
• Data used for training and testing has to
The use of data for the testing of AI systems shall testing and validation must be documented.
comply with applicable data protection, and privacy
be representative and generalizable. laws. • Transparency on how personal data is The methods used for developing, testing and
validating, and the outcomes of or decisions made
It shall be trained and tested on as expansive as processed must be provided. by the AI system shall be fully documented as part
is feasible, representative, relevant, accurate, and of the development process according to SAP’s
In alignment and compliance with applicable data
generalizable datasets. protection, and privacy laws, products that use AI Global Development Policy and Product Development
systems in the processing of personal data must Standards.
provide transparency to the extent possible as to how
the AI system was used in clear and simple language if
requested by the data subject.

04
Producti-
AI Use Case
Owner
AI Engineer Developer Designer zation
We will follow
& document
these requirements:
AI OPS User Assistant
A.
Engineer Developer
Human Agency & Oversight.

Example:
The developed AI functions are now ready to be embedded composed AI systems. be tested extensively to avoid unintended
into the product. Since the data sets in production will be When two or more AI systems are connected to each behaviour.
located somewhere else than the training data sets, the other or embedded within each other, additional testing When human-on-the-loop models are used,
corresponding documentation is refined and extended and control measures should be performed for the
appropriate extensive testing and governance shall
be conducted during development and deployment
by the AI engineer. Developers and UX designers work to ensure the system behaves as intended by the
on a user-friendly way to display local explanations for • Clear and simple explanations have to be developers and does not have any unintended
the predictions made by the AI directly in the application. provided for automated decisions. behaviour, outputs, or usage.
Furthermore, they include a feedback control in the As far as is practical, a clear and simple explanation
product, which allows users to share feedback on the AI system used in automated decision processes.
quality of the AI recommendations. This feedback could be
used to re-train the algorithm, if needed.

04
Producti-
• Unfair bias must be avoided. • Model architectures must not include

zation
SAP shall endeavour to achieve fairness; AI systems ethically objectionable features or
C.
be used. subject.
discrimination, or marginalization of groups or • The AI system should addresses the subjects upon request, describing the data segment
individuals, e.g. by reducing bias in training data. widest possible range of end-users. the subject was placed into and the reasons they were
comply with applicable data protection, and privacy

05
Operations
AI Use Case AI OPS User Assistant Customer We will follow
Owner Engineer Developer
& document
these requirements:
Example:
When the embedded AI function is deployed the AI Ops
A.
Engineer makes sure that the data used for on-boarding
the customer is representative and regularly checks
for undesired bias during execution. Further the AI Ops Human Agency & Oversight
Engineer provides potential user requests in regards to
transparency and explainability to the Use Case Owner.
composed AI systems. be tested extensively to avoid unintended
Because of the need for continues improvements by the behaviour.
When two or more AI systems are connected to each
AI Factory Process, one year after the successful product other or embedded within each other, additional testing When human-on-the-loop models are used,
launch, the feedback control shows that the accuracy and control measures should be performed for the appropriate extensive testing and governance shall
of the predictions provided by the AI by including two be conducted during development and deployment
to ensure the system behaves as intended by the
new parameters needs to improve. The AI OPS Engineer • Clear and simple explanations have to be developers and does not have any unintended
advises the development team to re-train the model provided for automated decisions. behaviour, outputs, or usage.
including the new parameters. Before the optimized As far as is practical, a clear and simple explanation
model is again built into the product, the team ensures AI system used in automated decision processes.
that the model is free from any biases.

05
Operations

• Model architectures must not include
ethically objectionable features or
C.
be used. subject.
discrimination, or marginalization of groups or • The AI system should address the widest subjects upon request, describing the data segment
individuals, e.g. by reducing bias in training data. possible range of end-users. the subject was placed into and the reasons they were
comply with applicable data protection and privacy

.
2023

AI Policy - EXTERNAL

Uploaded by

Copyright:

Available Formats

AI Policy - EXTERNAL

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AI Policy - EXTERNAL

Uploaded by

Copyright:

Available Formats

Development

SAP Definition of AI 0.6

Ethical AI Guiding Principles 0.8

The AI Ethics Policy 0.10

Red Lines 0.12

AI Factory Process 0.18

A Human Centered Approach 0.20

01. Ideation 0.24

© 2023 SAP SE or an SAP affiliate company.

Updated version: August 2023

2 SAP AI Ethics Handbook | External

4 SAP AI Ethics Handbook | External

• Rule-based AI systems are characterized by the fact that

6 SAP AI Ethics Handbook | External Development Chapter 7

8 SAP AI Ethics Handbook | External Development Chapter 9

01. 02. 03.

01. RED LINE

DPP-350 (Product standard)

10 SAP AI Ethics Handbook | External Development Chapter 11

Human Surveillance 02.

12 SAP AI Ethics Handbook | External Development Chapter 13

SAP High Risk Cases Determining High Risk Cases

Please be mindful that further changes could be Automated decision making:

Negatively affected individuals:

14 SAP AI Ethics Handbook | External Development Chapter 15

Risk Classification & Assessment Process

yes 01. 02. 03.

Red Line Case

Complete the AI Ethics

16 SAP AI Ethics Handbook | External Development Chapter 17

18 SAP AI Ethics Handbook | External Development Chapter 19

A In SAP we believe that a greater emphasis and under-

This enables that our products are as close as possible

20 SAP AI Ethics Handbook | External Development Chapter 21

Factory Process Activity:

AI Data Scientist AI OPS Engineer

Factory Process Activity: Factory Process Activity:

User Assistance Developer

Factory Process Activity:

Product Developer Customer

Factory Process Activity: Factory Process Activity:

22 SAP AI Ethics Handbook | External Development Chapter 23

24 SAP AI Ethics Handbook | External Development Chapter 25

• Affected users should be involved in the

• The AI system should address the widest

• Prevent misuse of the AI system in

26 SAP AI Ethics Handbook | External Development Chapter 27

28 SAP AI Ethics Handbook | External Development Chapter 29

• Unfair bias must be avoided.

• Inclusive data must be used for training.

30 SAP AI Ethics Handbook | External Development Chapter 31

C. Transparency and Explainability

• AI systems interacting directly with

32 SAP AI Ethics Handbook | External Development Chapter 33

Human Agency & Oversight

As far as is practical, a clear and simple explanation

34 SAP AI Ethics Handbook | External Development Chapter 35

• Unfair bias must be avoided.

36 SAP AI Ethics Handbook | External Development Chapter 37