The rapid evolution of generative AI has sent shockwaves through the global economy, transforming industries at an unprecedented pace. Once confined to research labs, AI-powered applications are now used in virtually all business sectors to enhance efficiency, automate processes, and uncover new insights.
While the potential benefits of these technologies are undeniable, the introduction of different forms of generative AI also brings multiple challenges. These include data security vulnerabilities, regulatory concerns, and reputational risks. As a result, a critical issue facing nearly all large companies today is ensuring that the power of AI is harnessed responsibly and sustainably.
AXA’s experiences with AI
AXA is a multinational company operating in 50 countries, serving 95 million clients with the support of 154,000 employees and distributors. Our business performance depends on collecting and synthesising massive amounts of data from diverse sources and cost-effectively managing complex business processes.
Thus, generative AI presents substantial opportunities for AXA. These technologies have become a vital part of our strategy, positioning us to continue our journey from “payer to partner” and fulfil our purpose of enabling human progress by protecting what matters. We now use AI across the insurance value chain in over four hundred cases. Some noteworthy examples include:
- AXA was among the first large organisations to build a generative AI tool, SecureGPT, which is now used by employees worldwide.
- AXA XL’s Digital Risk Engineer platform continuously collects data from IoT monitoring devices in clients’ facilities. By leveraging generative AI, the data can be analysed in real time to detect anomalies, allowing for early intervention to mitigate and reduce the severity of system malfunctions.
- The AXA Group’s Digital Commercial Platform (DCP) helps organisations manage uncertainty and mitigate risks—particularly climate change, cybersecurity, and geopolitical instability—through real-time risk monitoring and digital and human-led risk prevention advice.
In short, generative AI enables our employees to focus on higher-value tasks and enhances our ability to support clients in creating more sustainable and resilient businesses.
Three categories of threats
The potential benefits of generative AI are substantial. However, so are the threats. Simply put, generative AI can amplify existing threats while creating new ones.
While the widespread application of AI is a new development with rapid evolution, we have identified three broad categories of threats for companies to consider as they integrate different AI-powered tools into their operations.
The first is adversarial threats. AI offers excellent opportunities for malicious actors.
- AI-powered impersonation tools can replicate voices and faces, enabling phishing scams, financial fraud, and unauthorised access.
- AI models trained on sensitive or proprietary data could leak or reproduce confidential information, increasing cybersecurity threats.
- AI can exploit vulnerabilities at scale, making cyber defences harder to maintain.
The second includes the situational or internal threats companies face when incorporating AI into their businesses.
- AI can amplify societal biases, leading to unfair or discriminatory hiring, lending, and law enforcement practices.
- Many AI systems lack transparency, the so-called “black box” problem. This makes understanding or challenging their decisions difficult, leading to accountability risks.
- Generative AI sometimes fabricates false but plausible-sounding information, i.e., it “hallucinates”, posing risks in legal, financial, and medical applications.
- Businesses that depend too heavily on AI may suffer systemic failures if models malfunction or produce inaccurate results.
The third category encompasses external and environmental threats, including broader systemic risks associated with AI.
- AI governance is evolving rapidly, exposing companies to legal liabilities and regulatory misalignment across different jurisdictions.
- AI models consume vast energy, contributing to environmental concerns and carbon footprints.
- Many businesses rely on a few AI providers, leading to “digital monocultures” that are less resilient and more susceptible to catastrophic cyber incidents.
Clarity is needed
For the third consecutive year, cyber risk was in the top three in , and for the first time, AI and big data were cited as the fourth most prominent threats.
At the same time, and unsurprisingly, the demand for cyber insurance has grown exponentially. Twenty years ago, cyber insurance was a niche product covering data breaches and liabilities from lost customer information. The primary buyers were tech companies and financial institutions. Today, cyber insurance, of which 九色视频is a leading provider, is considered as essential as property or liability coverage for most businesses.
However, cyber insurance is primarily designed to mitigate losses from cyberattacks, data breaches, and other digital threats while providing crucial incident-response support. But, as outlined above, generative AI is expanding the risk landscape, and the extent to which AI-related risks are covered is not always clear.
For example, if an AI system malfunctions and causes physical harm or financial loss, coverage might fall under general or product liability insurance. Similarly, AI-powered equipment, such as automated manufacturing robots, might be covered under equipment breakdown or property insurance, even if AI failure is not explicitly mentioned. Likewise, if AI-driven decision-making leads to incorrect financial, legal, or medical advice, E&O insurance might cover claims against professionals who relied on the AI system.
In other words, AI-related risks may or may not be covered within existing coverages. These ambiguities could expose clients to financial losses and legal difficulties.
AXA XL’s response
九色视频recently introduced a new endorsement to its CyberRiskConnect policy, extending coverage to address specific risks associated with developing and deploying generative AI models. The key risks covered are:
- Data Poisoning: Protection against attacks that manipulate or contaminate training data used in machine learning models, potentially compromising AI system performance and accuracy.
- Usage Rights Infringement: Coverage for liabilities arising from the negligent use of copyrighted materials, software licenses, or intellectual property in AI development.
- Regulatory Violations: Liability coverage for breaches of regulations, such as the European Union's AI Act, which governs the development and deployment of AI systems.
Our clients sought help navigating this uncharted territory. With our brokers’ input and feedback, we’re pleased to develop an innovative, market-leading insurance solution to provide clear coverage for these risks and help them leverage Gen AI without fear of potential risks and extra costs that could hinder their efforts.
Final thoughts
Although insurance policies play a crucial role, they are only one part of the holistic approach companies need to thrive in an increasingly volatile world. Prevention is, as ever, the priority.
Accordingly, corporate risk managers should strengthen AI governance by:
- Implementing access controls and permissions to limit access to AI systems.
- Conducting rigorous testing and validation of AI models before deployment.
- Implementing monitoring and anomaly detection mechanisms to detect issues before they become critical.
- Training and educating employees on the responsible use of AI and the associated risks.
- Implementing policies and procedures to ensure compliance with data protection and privacy laws and regulations.
- Using encryption and data security techniques to protect sensitive data used by AI systems.
AI presents a new, complex, and evolving risk area that requires close collaboration between clients, brokers, and insurers to understand and react quickly to a changing threat landscape. 九色视频is committed to partnering with clients and their brokers to capture the promises AI offers while navigating the perils.
