What AI policies exist at AXA? What do they cover?

The Group AI Policy, supervised by our Group Technology and AI Officer, sets requirements for the consistent application of AI practices within AXA. From the identification of AI systems or prohibited AI practices to the AI risk assessment and mitigation, it establishes a clear AI governance and risk management process to ensure responsible AI development and usage.
The Group AI Compliance Policy, supervised by our Group Compliance Officer, outlines minimum requirements for compliance with the EU AI Act and any other applicable local AI regulation.
AXA entities develop their local AI policies, aligned with those applicable at Group level.

What are the AI do’s and don’ts at AXA?

At AXA, AI is used with the goal of creating value for our customers and people, such as providing better customer services, optimizing the claims management process, and improving call centers efficiency.
AXA’s AI policies clearly establish that AI cannot be developed or used for prohibited AI practices breaching applicable law (including the EU AI Act). This includes clear prohibitions regarding the exploitation of vulnerabilities, manipulative behaviors or social scoring.  In addition, we rely on our Responsible AI governance, risk management, guardrails and oversight mechanisms to prevent AI from operating outside its intended purpose.

How does human oversight look like at AXA?

At AXA, we believe that the required degree of human oversight varies based on the specific use case and the associated risk exposure. For example, based on our Generative AI Architecture Principles, generative AI applications should involve human supervision, intervention, and collaboration throughout the process. Additionally, we assess, determine, and implement the appropriate level of human supervision when integrating Generative AI into existing processes.

Will I know if I’m interacting with AI or using AI-generated content?

Yes. AXA will inform you whenever you directly interact with AI within an AXA environment, in accordance with our Transparency principle (one of the seven Responsible AI Principles). We also encourage employees to exercise critical thinking towards AI-generated content and to verify the generated content before use.

Can I enquire about an AI-driven decision or outcome and challenge it?

You can contest automated decisions, including those related to AI, by using the channels provided by AXA to raise feedback or disputes, such as writing to privacy@axa.com. AXA ensures that your feedback is properly taken into account, provides you with the information and means to express and resolve concerns, and ensures procedures to respond promptly to your enquiries, including those related to your data.

What measures have AXA put in place to reduce the environmental impact of AI?

At Group level, AXA AI Research team and Digital Sustainability teams have developed an open data model to estimate the carbon footprint of AI scenarios. This tool aims at supporting data-driven decisions to pilot AXA's operations sustainably with the aim of limiting the associated environmental impact.
At local level, AXA France, has implemented a framework for managing environmental impact of Generative AI use cases, estimating the carbon footprint and promoting eco-design.
Data scientists across AXA entities have received guidance for sustainable AI implementation. Regular communication and learning contents are established, encouraging sustainable practices among the users of our Generative AI tool (SecureGPT) across AXA. Our Data, AI and Digital Sustainability communities’ webinars and animations ensure that Sustainable AI initiatives are promoted across the organization and inspire others to replicate.