Case study: AstraZeneca featuring Margi Sheth, R&D Data Policy Director

Pharmaceutical giant AstraZeneca is making use of advances in AI technology. Current use cases include enhancing its understanding of diseases, matching patients with clinical trials and optimizing medicine supply chains.

To ensure the technology is used responsibly and effectively, the company has implemented five overarching ethical principles. It was the first global pharmaceutical business to publish these online.

They state that AI should be:

• private and secure
• explainable and transparent
• fair
• accountable
• human-centric and socially beneficial.

Taking fairness as an example, the company states that it endeavors to use robust, inclusive datasets in its data and AI systems and that it will treat people and communities fairly and equitably in the design, process and outcomes distribution of its AI systems. A more detailed set of objectives and principles for each pillar are published on its website.

“We want to make our ethical principles enabling rather than restrictive. Through training, we try to equip our teams with the tools and knowledge they need to do digitalization responsibly. It’s about encouraging the right behaviors and culture.”

Margi Sheth | R&D Data Policy Director, AstraZeneca

Determining these principles was not easy as there was no pre-existing guidance or draft regulation. At the time, major automotive companies and technology giants had published principles for ethical AI use. But AstraZeneca required something unique because some applications of its AI technology involve patient information or even treatment decisions, which are highly sensitive and require patient consent.

As part of the process of establishing its ethical principles, Sheth crowdsourced ideas from different areas of the business, including R&D, commercial, legal, business development and HR, as well as external academics, thought leaders and philosophers. Eversheds Sutherland also participated in this process.

“Ethics is not necessarily just the law, it’s about the right thing to do,” says Sheth. “But what is the right thing to do? You can only form a view by engaging across the entire business and also externally.”

Converting principles into practice

Establishing ethical principles was just the first step. Sheth and her internal and external team then implemented a suite of governance, training, guidance and escalation procedures to ensure these principles were implemented across the business.

A key aspect of this is partner due diligence. “If we want to work with someone outside, we want to encourage them to live by the same values as us,” says Sheth. “And if there’s a discrepancy in values, a partnership is never going to be the best it can be.”

As a result, teams that forge alliances with AI businesses or those that heavily use AI are encouraged to consider ethical issues in addition to commercial and legal factors. Therefore, depending on the circumstances, it may ask potential partners how its AI was developed and how transparent they are as part of wider due diligence.

To aid this process, AstraZeneca - in collaboration with Eversheds Sutherland – developed cutting-edge fit for purpose AI contracting templates and guidance that can be used for procuring technology and partnering with AI businesses. They focus on the following key areas:

• ensuring shared ethics in key areas
• transparency and auditability
• clear rules including around fair use of data, rights to use data, training of data, non-discrimination and bias 
• environmental and sustainability considerations
• clear provisions around ownership of input and output data and models

Ownership provisions are particularly important given the changing IP laws in this area. The templates also address other key technology and procurement considerations including cybersecurity, privacy considerations, warranties and liability, service levels and more. Crucially, the teams that use the templates continue to receive training on the AI legal and ethical landscape and on how contract templates can be used with suppliers in practice.

Ensuring uptake

Sheth is keen for the entire business to buy into its ethical Data & AI principles. As a result, it has adopted a collaborative approach, educating teams about the benefits. “Through training, we try to equip our teams with the tools and knowledge they need to do digitalization responsibly. It’s about encouraging the right behaviors and culture rather than imposing rigidity.”

The business is keen that any measures introduced to ensure its ethical procedures are adhered to but do not slow innovation. One tactic it has used is to leverage existing governance review processes. “As our data security review processes are updated, we incorporate our AI privacy and security questions and approach at the same time, rather than creating a completely new form and approach. It’s all about making the process as user friendly as possible.”

As AI continues to transform the entire organization, Sheth anticipates that the methods to develop and deploy AI responsibly will continue to evolve. Future plans include general employee training on AI ethics and cementing a culture of responsible AI use.