Integration of AI solutions introduces cybersecurity risks including adversarial attacks and unauthorized employee use.
“AI solutions may use algorithms, datasets or training methodologies that are incomplete, reflect biases, or contain other flaws or deficiencies. AI solutions may create output that appears correct but is inaccurate, biased or otherwise flawed, or that infringes or otherwise violates intellectual property or other rights. Our integration of AI solutions internally for business purposes and into our offerings also introduces unique cybersecurity risks, including gaining unauthorized access to our products, adversarial attacks intended to deceive or 'poison' AI models, and model extraction attempts to steal our proprietary algorithms. We also face risks of employees using unauthorized AI tools that have not been vetted and approved by us and inadvertently disclosing confidential information to such third-party AI platforms. As AI-driven threats become more automated and sophisticated, they may bypass traditional security controls at a speed and scale that exceeds our ability to respond, potentially leading to significant operational disruption or the loss of sensitive intellectual property. We may not be able to control the development, maintenance or behavior of third-party AI solutions, including the autonomous actions of AI agents or the security of third-party foundational models, and these AI solutions may be used inappropriately or introduce novel vulnerabilities into our software via AI-generated code. We, and third-party providers of any AI solutions we may make availabl”