At Rise of AI 2024 in Berlin, Dr. Sarah J. Becker, Partner at Deloitte, argues that AI ethics is a strategic enabler rather than a compliance afterthought. Opening with a personal story about a panic attack on a flight, she describes how people accept transformative technologies without questioning them until something forces reflection, and how new technologies tend to produce polarized reactions, either overestimating their potential or fearing humans will be overrun. She presents survey data showing that 79% of decision-makers already using generative AI report ethical concerns, yet only one in five companies has an ethical framework in place. Drawing on a study of AI ethics guidelines, she shows transparency as the most cited value worldwide, with explainability more prominent in Europe and fairness more prominent in East Asia. She warns that many organizations are currently swinging from early AI experimentation toward heavy control through compliance, legal and security functions, which raises the barrier to innovation. Her closing message is that ethics cannot be delegated and needs real resources, not just guideline documents.
Key takeaways
- 79% of decision-makers already using generative AI report ethical concerns, but only one in five companies has an ethical framework in place.
- Becker distinguishes visible concerns like hallucination and data privacy from a deeper iceberg of fears about jobs, inequality and loss of control.
- Her guideline study finds transparency the top AI ethics value worldwide, present in 93% of the European guidelines studied.
- Explainability is comparatively more emphasized in European guidelines, while fairness is comparatively more emphasized in East Asia, notably China.
- She warns companies are currently swinging from AI experimentation toward compliance-driven control, which raises the barrier to innovation.
Detailed notes
Why new technologies produce polarized reactions
Becker opens with a personal anecdote about having a panic attack on a flight to Boston in 2016, which made her realize she had never rationally questioned the technologies that let her live as a frequent traveler. She uses this to argue that when a new technology has the power to transform how people live, reactions tend to polarize between those who overestimate its potential and those who fear humans will be overrun. She states that AI is one of those technologies and is here to stay, so organizations do not have the option of staying in a polarized stance; they have to build strategies that balance innovation and control, which she says is harder in practice than it sounds.
The ethics gap: survey data and the iceberg of concerns
She cites a finding that 79% of decision-makers in companies already using generative AI report having ethical concerns, while only one in five companies has an ethical framework in place to manage them. The most commonly cited concerns are hallucination, bias and data privacy, but a deeper study across an international company uncovers a larger 'iceberg' of concerns: fear of job displacement, fear of growing inequality, racism, surveillance and loss of human control over machine decision-making. She argues hallucination is largely a technical, solvable problem tied to data quality and trustworthy AI, while data privacy and intellectual property are often regulatory rather than purely ethical questions, depending on the industry.
Which values guide AI ethics, in Europe and globally
Drawing on a study of AI and digital ethics guidelines, Becker reports that transparency appears in 93% of European guidelines, alongside beneficence, explainability, autonomy, responsibility and security, values she notes also underpin the pillars of the EU AI Act. Comparing Europe with two other global regions, she finds transparency, accountability and security consistently at the top everywhere, but explainability is comparatively more emphasized in Europe, while fairness is comparatively more emphasized in East Asia, particularly China, where she links it to acceptance of social scoring systems. She warns that shared top-line values can mask real differences in what each value actually means in context.
From guideline word to working practice
Using beneficence as an example, Becker describes how a value moves from a dictionary definition into a corporate setting: first a stated commitment to the value, then measures and metrics, supported processes and KPIs, and people made responsible for it, so that employees doing the actual work understand what the value means for their tools, questions and products, not just a statement on a corporate website. She observes that companies are currently swinging toward control, with cybersecurity, compliance and legal functions taking over AI governance after initial exploration phases, which raises the barrier for innovation even as the pressure to keep pace continues.
Closing message
Becker concludes that ethics cannot be delegated: leaders can assign responsibility for it, but if it is sidelined it becomes a barrier to successful transformation. She calls on organizations to give ethics real resources and budget, not just a symbolic seat at the table, arguing that companies which invest seriously in an ethical tech mindset are further along in their AI transformation than those that have not, and that this requires bridging accountable leadership with the people who create value day to day.
Quotes
Ethics cannot be delegated.
Sarah J. Becker
AI is one of those technologies and it is here to stay.
Sarah J. Becker
Give ethics a seat on a table.
Sarah J. Becker
Questions answered in this talk
How many companies have an ethical framework for AI?
According to Dr. Sarah J. Becker at Rise of AI 2024, 79% of decision-makers in companies already using generative AI report having ethical concerns, but only one in five of those companies has an ethical framework in place to manage them.
What is the difference between visible and hidden AI ethics concerns?
According to Dr. Sarah J. Becker at Rise of AI 2024, the most cited concerns are hallucination, bias and data privacy, but deeper research reveals a larger set of hidden concerns including fear of job displacement, inequality, racism, surveillance and loss of human control over machines.
Which AI ethics value is most emphasized around the world?
According to Dr. Sarah J. Becker at Rise of AI 2024, transparency is the most cited value across AI ethics guidelines worldwide, appearing in 93% of the European guidelines her study reviewed, with explainability more prominent in Europe and fairness more prominent in East Asia.