For enterprise leaders and their legal & compliance advisors Ethical Considerations for the Use of AI

1 Contents

Exploring the ethical use of AI 3 The AI dilemma 3 Defining your ‘north star’ for AI 4

What is AI? 5 Defining AI 5 What makes AI possible? 6 Some examples of AI 6 Reaching human parity 7 The economic impact of AI 8 The promise of AI 9

Microsoft’s approach to AI 10 An overview of ’s AI solutions 10

But what about AI ethics? 11 AI must be built on an ethical foundation 11

Microsoft’s framework for ethical AI 12 Microsoft’s six AI principles 12

Recommendations for how to approach AI Ethics in your organization 18 AI skills are the key 18 Facial recognition technology: The need for public regulation and corporate responsibility 19

Ensuring AI empowers us all 20 Using AI to build a better world 20

For more information about AI and GDPR 21

Ethical Considerations for the Use of AI Exploring the ethical use of AI

As with all great technology advances, artificial intelligence will bring about vast changes some of which are hard to imagine today. How can we be sure the actions taken by an Artificial Intelligence or AI system are actions that are in our best interests?

The AI dilemma

Microsoft’s corporate mission is “to empower every our AI systems are making the right choices and person and every organization on the planet to achieve decisions? How can we be sure the actions taken by more.” In the past, this mission has been tackled using an AI system are actions that are in our best interests? traditional computers, mobile devices, software, How can we be sure that an AI system doesn’t make connectivity and more recently with cloud computing. decisions or take actions that humans would consider But now, we are witnessing a technological revolution unfair or unethical? as platforms and machines become more intelligent than ever before, opening up new ways to empower humans and organizations.

These advances will make systems easier to use, bring more data to our fingertips and create new ways to address every day needs. New intelligent systems— working with humans—have the potential to also solve some of society’s greatest challenges such as poverty, disease and lack of education. Together, these types of In order to answer these questions and address these new smart technologies have been defined under the concerns, we need to remember that AI systems are term “artificial intelligence” or AI. ultimately built by humans. These systems work by using the data we provide and the algorithms we As with all great technology advances, artificial create that leverage the data to perform a desired intelligence will bring about vast changes some of function. It is up to us as designers, programmers, which are hard to imagine today. As we develop lawyers and managers to ensure the systems are computing systems to be increasingly intelligent we built to behave in way that is consistent with our own are all faced with a dilemma. We want these systems beliefs, laws and ethical standards. to be useful and truly intelligent yet how do we know

3 © 2018 Microsoft Corporation. All rights reserved. Defining your ‘north star’ for AI

In 2016, Microsoft’s CEO published a transportation, agriculture and education while set of principles on the ethical use of AI. As Microsoft exploring some of the complex questions and broad continues to develop AI-based solutions, it remains societal concerns that these advances will bring about. focused on ensuring that its AI solutions and AI tools For an in-depth discussion of AI and the ethical are built on a foundation of these ethical principles. principles discussed in this paper as well as the impact In January 2018, Microsoft published the book of AI on our future, we encourage readers to read The The Future Computed which highlights the potential Future Computed. for AI to augment the abilities of humans to create breakthrough advances in areas such as healthcare,

This paper presents some of Microsoft’s ideas on how organizations can understand both the benefits and the impacts of AI and provides some recommendations for how to ethically deploy AI solutions in a fair and transparent way.

4 What is AI?

Artificial Intelligence or AI has become a widely used, catch-all phrase that pops up in technical discussions, science fiction, analyst reports, news and beyond. It seems to get applied to almost any technology that is ground breaking or futuristic. But most experts agree that AI can be precisely defined in terms that even non- technologists can understand.

Defining AI

Perception, learning, reasoning and decision- making are key features of human intelligence. When computers perform these operations, they are said to be performing Artificial Intelligence (AI) routines. Recent advances in AI have been fueled by improvements in machine learning techniques that can identify patterns in large data sets and use these patterns to make predictions or recommendations. The systems learn from a part of the data, their predictions are then tested against the remainder of For more information on AI, be sure to read our AI the data. They are further refined to be increasingly Insights feature: A Short Introduction to Artificial more accurate. It is the learning component that Intelligence. makes AI different from historical approaches to building machines or programming software. AI is not explicitly programmed to respond a certain way, it learns to respond that way.

5 © 2018 Microsoft Corporation. All rights reserved. What makes AI possible?

The recent advances in AI to understand and respond AI development is at an inflection point, where the to the world are underpinned by two important vast amounts of available data, combined with Internet trends made possible by the development of cloud connectivity and affordable cloud computing power, computing—the ability to collect and store huge have facilitated substantial advances in innovative amounts of data and the ability to assign vast compute algorithms in the last 15 years. The promise of AI is power to analyze this data. that the knowledge gained from applying analytics and machine learning to the available data will enhance any decision-making process with additional insights and intelligence, leading to better outcomes and improvements in every aspect of people’s lives and the world around us. A better term for AI is “computational intelligence” where the advantages of computer processes are harnessed to help people make better decisions.

Some examples of AI

AI solutions are also pushing the boundaries of enabling people to interact with systems more naturally. For example:

• Vision: the ability to recognize a picture or a video.

• Speech: the ability to listen and transcribe words in a text.

• Language: the ability to comprehend a variety of languages, evaluate sentiment and learn how to recognize what users want.

• Knowledge: the ability to detect and draw inferences on how disparate pieces of information are related to each other. For example, AI systems can predict how bad traffic will be, based on historical traffic flows for the time of day and season, while also looking at real-time data from weather conditions, incidents and events in the area.

Computers are learning the way people do, through experience, and for computers, experience is captured in the form of data. Fundamentally, AI is about computers understanding and responding to events in the world. There are two pieces to this—perception and cognition.

6 For example, “perception” means the system can computers have also gained cognition or the ability to recognize faces, identify objects in images, transcribe learn and reason. speech, classify emotions, understand simple commands, or diagnose a disease. Computers are now able to comprehend and translate languages and to study patterns and make predictions. “Cognition” goes further by analyzing data to translate sentences, answer questions, drive cars, play chess, determine when a part needs repair or recommend a medical treatment.

Humans can perceive and understand, and now computers are increasingly able to replicate these abilities in many areas. Machines have been able to see and hear for a long time. Cameras have been capturing images since 1839 and sound recordings have been possible since 1877. Modern computational machines not only capture images and sounds. They are now Skype Translator allows students and teachers to increasingly able to identify what they are seeing connect over Skype and chat via a video call with live and hearing. Beyond identifying images and sounds, translation.

Reaching human parity

In many ways, modern AI systems have already come close to meeting human capabilities in several key areas. For example, has developed AI technology that has achieved some remarkable milestones:

• 96% of image recognition parity with humans—as good as a college graduate.

• 95% of speech recognition parity with humans—the standard of professionaltranscriptionists

• Strong progress in translation— can translate 62 languages

• Almost 90% reading parity with humans

7 © 2018 Microsoft Corporation. All rights reserved. Computers are learning the way people do, through experience, and for computers, experience is captured in the form of data. Fundamentally, AI is about computers understanding and responding to events in the world. There are two pieces to this— perception and cognition.

And yet, while AI systems are reaching parity in these discrete tasks, these systems will be challenged to deal with unique human concepts such as empathy and emotion.It is also worth noting that none of these areas have reached 100% parity. But humans don’t operate at 100% either and make errors all the time. AI doesn’t need to be 100% accurate to be valuable.

The economic impact of AI

Industry analysts, governments, and Microsoft believe production that “promises to transform the basis of that AI can positively transform both developed and economic growth for countries across the world.” developing economies. This chart shows Accenture’s estimate of the positive A recent study by Accenture argues very persuasively growth-accelerating effect AI is likely to have on that “artificial intelligence is the future of growth.” some of the world’s most advanced economies. The According to this study, increases in capital and light blue bars indicate Accenture’s assessment of labor are no longer driving the kind of sustained each economy’s baseline growth potential, while the and truly global economic growth we’ve seen in the dark blue bars show its forecast for these economies past. But Accenture’s study of AI’s potential has led boosted by AI. it to conclude that AI is becoming a new factor of

8 The promise of AI

Microsoft, in line with its corporate mission “to capabilities, including speech and image recognition empower every person and every organization on to help the visually impaired to “see” the world around the planet to achieve more,” firmly believes in using them. AI to achieve these goals, leading to advances in nearly every field of human endeavor. When made Environmental sustainability accessible to everyone, AI will transform industries, AI can also enable us to better address environmental make us more productive, help solve the world’s concerns. Microsoft launched AI for Earth in London biggest challenges, and lead to important advances in in June 2017—a new initiative dedicated to the education, healthcare, transportation, agriculture and sustainability challenges of agriculture, water many other areas. management and distribution, biodiversity, and climate change. Healthcare For example, in healthcare, AI systems are already With the world’s population expected to grow helping people tackle complex problems. A good by nearly 2.5 billion people over the next quarter example is Project InnerEye, a project in which U.K.- century, AI offers significant opportunities to based researchers at Microsoft have teamed up with increase food production by improving agricultural oncologists to develop an AI system to help treat yield and reducing waste. For example, in precision cancer more effectively. farming, sensors are used to provide real time measurements such as moisture and soil conditions, As another example, AI can reduce hospital which are then analyzed to help provide farmers with readmission, enhance the quality of care for managing recommendations on how to maximize their yields. chronic disorders, and keep hospitals safe and efficient.

An Institute of Medicine study released in May 2016 estimated that preventable errors in hospitals are the third leading cause of death in the US, trailing only heart disease and cancer. The number of deaths is estimated to exceed 250,000 patients per year. AI systems can be developed to catch errors by recognizing anomalies in best clinical practices, saving thousands of patients per year.

Overcoming disabilities AI for Earth puts Microsoft cloud and AI tools in the hands Another area where AI has the potential to have a of those working to solve global environmental challenges. significant positive impact is in serving the more than Through grants that provide access to cloud and AI tools, one billion people in the world with disabilities. One opportunities for education and training on AI, and example of this is Microsoft’s “Seeing AI” application. investments in innovative, scalable solutions, AI for Earth Saqib Shaikh, a Microsoft senior software engineer works to advance sustainability across the globe. in the UK, lost his sight when he wasseven years old. He was inspired to develop an application that uses AI

9 © 2018 Microsoft Corporation. All rights reserved. Microsoft’s approach to AI

Microsoft is both a supplier of AI technology and company that increasingly uses AI to run its business. In both cases, Microsoft believes that AI must be used to benefit employees, partners, customers and society as a whole.

An overview of Microsoft’s AI solutions

AI offers incredible opportunities to drive widespread Microsoft is also infusing AI into its end-user products. economic and societal progress and Microsoft’s For example, Windows is more secure thanks to mission is simply to make the computational AI systems that detect malware and automatically intelligence of AI available to everyone. Microsoft protect computers against it. PowerPoint Presentation is using AI capabilities to develop and enhance its Translator enables auto-captioning and translation products and services (such as Windows and Office), into over 60 languages. The auto-captioning feature as well as making AI tools available to developers and can also help people with hearing loss to engage more organizations to build their own AI-powered products. productively at work.

The Microsoft AI platform offers services, tools, and And Microsoft’s AI business solutions include infrastructure to make AI development easier for intelligence systems to improve an organization’s developers and organizations of all sizes. Microsoft’s ability to gain insights and formulate better courses of service offerings include Microsoft Cognitive action using the information that they have gathered. Services—a set of pre-built AI capabilities, including vision, speech, language, search, and knowledge capabilities, that are run in the cloud and can be easily integrated into applications with just a few lines of code. Other technologies help simplify the creation and integration of “bots” that can improve the customer service experience. A set of infrastructure offerings include tools that make the AI development process easier.

Microsoft also Open Sourced the Microsoft Cognitive Toolkit, a set of capabilities used by Microsoft engineers to achieve breakthroughs in AI and is contributing data sets to advance development of machine comprehension of text.

10 But what about AI ethics?

The introduction of every new technology throughout history has brought about both positive and negative changes with regard to humans and the work they do. With AI systems becoming capable of cognitive tasks once only thought possible by humans, how do we ensure that AI only helps humans and never harms us? As computers are able to interact with the world in the same way that humans do, how will those interac- tions impact real people?

AI must be built on an ethical foundation

We are now empowering computers to use AI tomake decisions without human input that werepreviously made by humans, or at least guided by human decisions. If we allow computers to perceive, learn, reason and make decisionswithout human oversight, it is important to think about the impact AI will have on people and the ethics that should underpin AI. Making the technologies available alone is not sufficient. For AI to be adopted and deployed on a global scale, we need to develop AI in a way that people believe to be trustworthy.

There is broad consensus about what kinds of behavior are considered responsible and acceptable for human behavior. In general, this consensus grows out of norms that are implicitly understood, consistently applied in everyday social interactions, and reinforced in laws. The same standards should apply to non-human, AI behavior. Making AI trustworthy can be realized only if relevant stakeholders from business, government, civil society, and the research community collaborate on shared principles and ethical frameworks. For AI to be adopted and deployed on a global scale, we need to develop AI in a way that people believe to be trustworthy.

Making AI trustworthy can be realized only if relevant stakeholders from business, government, civil society, and the research community collaborate on shared principles and ethical frameworks.

11 © 2018 Microsoft Corporation. All rights reserved. Microsoft’s framework for ethical AI

At Microsoft, over the last couple of years there have been discussions at every level within the company and from every discipline about how to build ethical AI solutions. These discussions have led Microsoft to believe that the development and deployment of AI must be guided by an ethical framework.

Microsoft’s six AI principles

As laid out in The Future Computed, Microsoft has established six core principles that we believe guide development of AI products and solutions. These consist of four core principles: fairness, reliability and safety, privacy and security, and inclusiveness—which are built on two foundational principles: transparency and accountability. We encourage others to consider how to apply these principles to their organizations, and how their own values may lead to development of additional principles.

12 Fairness At the most fundamental level, AI systems should treat identify different perspectives that need to be everyone fairly and impartially and not affect similarly considered during the development process. situated groups of people in different ways.We all need to ensure that AI is not programmed to make biased or When AI-based recommendations are applied, a full discriminatory decisions. understanding of the limitations of the system is essential to ensure that the outcome does not lead to The design of any AI system starts with the choice of unequal impacts. training data—which is also the first place where bias can enter the system. Training data should sufficiently For example, a credit rating algorithm that suggests a represent the world in which we live. AI researchers person has an 80% probability of being a credit risk is have discovered, to their chagrin, that when they are also suggesting that, in 20% of the cases, someone with not careful in training AI algorithms in scenarios like similar characteristics is not a risk. This insight needs face recognition, these algorithms can perform poorly to be taken into consideration by people who are using on groups that have only a small representation in the the algorithm to make decisions that impacts people. training data. These types of systems should be used inform people who ultimately still need to be responsible for the final For example, some early facial recognition algorithms decision. were trained using data sets that consisted mostly of photos of white men. It was then discovered that these It is essential that AI does not limit opportunities for algorithms didn’t do a good job at recognizing women anyone—fairness is the foundation for treating people or people of color. The solution to this problem is to with dignity and respect. make sure that the data used to train AI algorithms completely captures the diversity of the populations the solution is dealing with.

If AI systems provide guidance on, for example, medical treatment, loan applications or employment they should make the same recommendations to everyone with similar symptoms, financial circumstances or professional qualifications. When developing these systems, developers should be sensitive to places where societal bias or other biases may be incorporated into either the training data or the algorithms.However, once these types of biases are detected, how to determine appropriate actions, Microsoft is committed to improving facial recognition corrections and responses, if any, that need to be taken so that it performs well across all skin tones and genders. is a policy issue that will require further dialogue with See Microsoft’s blog article on this here. the different stakeholders involved.

As AI systems are ultimately designed by people, attracting a diverse pool of AI talent will help to

13 © 2018 Microsoft Corporation. All rights reserved. Reliability and safety We must ensure that AI technology is reliable and safe, • Expect the unexpected—including nefarious as prior new technologies were made reliable and safe attacks as they became readily available. The complexity of AI technologies has fueled fears that AI systems may Moreover, human oversight and/or intervention in cause harm in the face of unforeseen circumstances or the execution and deployment of AI systems will be that they can be manipulated to act in harmful ways. critical. However, the human-system interaction must Trust will then depend on whether AI systems can be be designed in such a way as to enable a person to operated reliably, safely and consistently even under assume control with sufficient information to make unexpected conditions, especially for applications in meaningful decisions. fields such as transportation, health care, and financial Microsoft is committed to working with the industry services—where consequential decisions are involved. to establish risk assessment standards for AI and These systems need to demonstrate that they expect developments in this area in the coming year. can be depended upon to make appropriate recommendations in as many circumstances as possible. Designing appropriate AI behavior in a very broad set of circumstances raises many challenging issues that are actively being researched.

Potential best practices to minimize any unintended consequences of actions taken by AI systems include the following:

• Evaluate training data for quality, suitability, and completeness

• Test extensively

• Monitor ongoing performance to detect potential Human oversight or intervention in the execution and issues when circumstances stray outside of deployment of AI systems will be critical. acceptable boundaries

Privacy and security Data Protection Regulation, on the collection, use, Perhaps one of the biggest areas of concern related and storage of data.Microsoft has invested in robust to AI is the need for privacy and security. Because the compliance technology and processes to ensure that foundation of AI is data, and AI systems that involve data collected and used by its AI system is handled human activities such as healthcare and education will legally and responsibly. need a lot of data about people, it is important that such personal data be securely stored, used safely and AI systems must be designed to protect personal to a good end or people will not share their data. data—regardless of where it is stored—from bad actors who want to steal private information or inflict It is also essential that AI systems comply with other harms. applicable privacy laws, such as Europe’s General

14 Best practices should be developed and implemented for tracking how personal data is accessed and used, and for enabling oversight of such access and use. There should be clear policies and transparency about data collection and use, and good controls so people can make choices about how their data is used. Microsoft leverages anonymization, strong encryption and other techniques to enable the data analytics necessary for AI while preserving individual privacy and security of data.

For more information on deploying AI and complying with Europe’s GDPR, be sure to read our AI Insights feature: Artificial Intelligence and the GDPR Challenge

For additional information on AI security, see Microsoft’s paper Securing the Future of Artificial Intelligence and Machine Learning at Microsoft.

Inclusiveness Enhancing the way that people interact with AI AI must be developed and deployed in a way that can systems through the use of contextually appropriate benefit all and is accessible by all. Microsoft’s vision and emotionally sensitive interaction design can of democratizing technology also applies to AI. In greatly enhance people’s trust in the system. For fact, Microsoft has invested $25 million in its AI for example, if this is the first time a person is using an Accessibility programs that harness the power of AI to AI-based personal agent, the agent can build greater amplify human capability for the more than one billion trust with techniques such as confirming the person’s people around the world with a disability. intentions and preferences (i.e., offer emotional intelligence), as well as making recommendations These programs and technologies can be powerful that are contextually appropriate (i.e., offer cognitive tools for people who are disabled (e.g., visually, intelligence). hearing, mobility, speech, or cognitively impaired), enabling them to gain access to opportunities in AI systems should benefit everyone and address education, employment, and government services a broad range of human needs and experience, previously unavailable to them, while also improving inclusively. their overall health, social/economic situations, and quality of life.

15 © 2018 Microsoft Corporation. All rights reserved. AI must be developed and deployed in a way that can benefit all and is accessible by all.

Transparency These first four core principles are underpinned by AI transparency. Microsoft is working with the two foundational principles.The first is transparency. Partnership on AI and others to develop best practices People can’t identify whether progress is being made for addressing meaningful transparency of AI systems. on the first four principles unless there is enough transparency around how systems have been built and function.

As AI systems are increasingly involved in decisions that influence people’s lives, providing clear explanations about how these systems operate as well as mechanisms for accountability are essential to building trust.

With many of the advanced machine learning techniques such as deep neural networks functioning essentially as “black box” systems even data scientists can’t fully understand them. To address this, researchers at Microsoft are developing a software toolkit that will enable developers to experiment with a variety of methods for explaining models and Should there be a Hippocratic Oath for AI? systems. This toolkit implements a number of simple In the foreword of The Future Computed the question models, including generalized additive models, and was posed whether we could see a Hippocratic Oath for several methods for generating explanations of the coders like we see for doctors that says that first they behavior of potentially complex models or their will ‘do no harm.’ individual outputs or predictions.

This is an area that will require further research to develop a holistic approach for addressing

16 Accountability The final foundational principle is accountability. This is perhaps the most important of all the principles since people who design and deploy AI systems must be accountable for how their systems operate.

Ultimately one of the biggest questions for our generation, as the first generation that is bringing AI to society, is how do we ensure that computers will remain accountable to people and how do we ensure that the people that design computers remain accountable to everyone else?

To establish norms and best practices, we can draw upon experience in other sectors such as healthcare and privacy.Internal review boards can provide guidance on which practices should be adopted during How do we to ensure that computers will remain development and deployment of AI systems. accountable to people and how do we ensure that the people that design computers remain accountable to everyone else?

17 © 2018 Microsoft Corporation. All rights reserved. Recommendations for how to approach AI Ethics in your organization

Microsoft believes the most important recommendation for any organization deploying an AI-based solution is to build up AI ethical expertise across every team in the organization.

AI skills are the key

The skills needed to acquire this expertise are a For example, Brad Smith, Microsoft’s President and combination of knowledge of relevant laws and Chief Legal Officer, recently issued a strong public call regulations, knowledge of the latest advances in what for government regulation related to the uses of facial AI can do, and knowledge of how to apply ethical recognition AI. Facial recognition has the potential principles to AI. Expertise is essential, and it takes to be used for very positive scenarios such as device dedication and focus for an enterprise ethics team to authentication, tagging friends in pictures, and new build the expertise it needs. With the appropriate skills capabilities for people with sight disabilities. in place, it may make sense to create a review board for AI ethics if AI is used by your organization in ways Yet, in order for facial recognition to work, it needs that touch people’s lives. data about your face, and such data should be processed in a trustworthy and compliant way. Plus, as Given AI is still at a nascent stage of development it mentioned earlier, these systems use the data that is is likely that changes to existing laws will be required available to them, applying algorithms that have been and new laws may be needed to keep AI on a path that created by humans. It is essential to ensure that these benefits society. While most companies do not get systems are used in ways that are not biased against or directly involved in working with policy makers, it is disadvantage any segment of the population. important to be aware of these potential changes and get involved where possible in the public dialog on AI advancements and legislation

18 Facial recognition technology: The need for public regulation and corporate responsibility

On December 6, 2018, Microsoft President, Brad Smith, outlined recommendations for legislation governing facial recognition technology and detailed progress and principles Microsoft has developed since he called on the need to regulate the technology back in July 2018.

In a speech to the Brookings Institutionin Washington, D.C, Brad proposed specific recommendations for policymakers to address this issue, and covered the challenges of facial recognition, especially around bias and discrimination, protecting people’s privacy, and protecting democratic rights and freedoms. He also announced six principles that Microsoft will adopt to govern the design, development and deployment of our own facial recognition technology.

“We believe it’s important for governments in 2019 to start adopting laws to regulate this technology. The facial recognition genie, so to speak, is just emerging from the bottle. Unless we act, we risk waking up five years from now to find that facial recognition services have spread in ways that exacerbate societal issues. By that time, these challenges will be much more difficult to bottle up.” – Brad Smith, Microsoft’s President and Chief Legal Officer

19 © 2018 Microsoft Corporation. All rights reserved. Ensuring AI empowers us all

The bottom line is an organization’s guiding principles or mission should be the foundation for the deployment of any technology or service. Microsoft believes in the case of AI, we must ensure that AI solutions are programmed to operate within those principles and help—not harm—society.

Using AI to build a better world

In closing, Microsoft believes in the potential of AI to transform the world and enable us all to address complex societal and economic challenges. But in order to do that, we need to continue to work together to not only realize the potential of AI but ensure that AI is used in an ethical, fair, and transparent way. We all need to ensure that AI technology, and the organizations that develop and use it, are governed by the rule of law and a strong ethical foundation.

For more information on this subject, we invite readers to explore the documents and resources mentioned in the next section and take proactive steps to get informed and ensure AI solutions are used ethically and responsibly in your organizations.

20 For more information about AI and GDPR

The Future Computed: Artificial and formulate best practices on AI technologies, Intelligence and Its Role in Society to advance the public’s understanding of AI, and to serve as an open platform for discussion and A short non-technical book discussing AI’s promise engagement about AI and its influences on people and challenge for modern society, by Microsoft’s and society. Link President Brad Smith and top AI researcher Link Microsoft AI A more technical introduction to Microsoft’s AI Facial Recognition: It’s Time for platform. Link Action The AI blog Microsoft President Brad Smith makes the case for The latest news about AI progress from Microsoft. careful regulation of AI-based facial recognition Link and an ethical approach by governments and corporations to all AI technologies. Link AI Insight Series: A Short Introduction • targeted online advertising to Artificial Intelligence Digital Transformation in the Cloud: • directWhat marketing enterprise actions leaders such as andemails their sent to An introduction to AI for legal and compliance professionals. Link selectedcustomerslegal and compliance or prospects advisors need to know • customizedA non-technical price or book product that lays offers out the made business to Artificial Intelligence—The Ethical specificcase forcustomers digital transformation based on their and explainsprofile the and Legal Implications foundational requirements of security, privacy, and A portal with articles and resources for legal and • evaluationcompliance. of individual Link online applications for compliance professionals by Norton Rose Fulbright employment, loans, insurance contracts, health related to ethics and AI. Link A Cloud for Global Good benefits, or university admission IAPP webcast: Artificial Intelligence, Microsoft’s policy considerations and Machine Learning, and Data Ethics • certainrecommendations kinds of scientific for creating research a framework or medical of laws diagnosisthat extend that involvethe benefits the analysisof the cloud of tonon- all. Link Hear three experts with legal, technological and ethical perspectives wrestle with the challenge of anonymized personal data (such as radiology ‘Transformation Tuesday’ blog series Artificial Intelligence, machine learning and data images or personal health records). ethics issues. Link A series of blog posts by Michael McLoughlin— Microsoft on digital transformation, AI, and cloud computing. Link

Partnership on AI The Partnership on AI was established to study

For further information, contact: Michael McLoughlin Director, Public Affairs Microsoft Corporation