When a transformative technology like generative AI emerges, it’s crucial for a company to adopt a clear stance. Generative AI has the potential to reshape how employees work and innovate, and it’s likely already in use within your organization—perhaps without your awareness. While the benefits are numerous, generative AI also comes with risks. By implementing a policy to govern its use, you can maximize its potential while minimizing potential downsides.
Intellectual Property (IP) Violations
The way generative AI models are built contributes to this risk. These models, trained on millions of images, music tracks, videos, and text, can create entirely unique content. However, there’s a risk that significant elements of the training data might appear in the generated content, leading to potential intellectual property issues.
The debate about whether AI-generated content infringes on the rights of original creators is ongoing. Some organizations advocate for creators to give explicit consent before their works are used to train AI models. Meanwhile, certain software companies are developing “commercially safe” generative AI tools trained exclusively on fully licensed content. Whether AI-generated content can be copyrighted varies by jurisdiction, but this question is crucial for businesses intending to use such content in external communications.
A generative AI policy can help mitigate IP risks by, for example, stipulating that AI-generated content cannot be used externally without thorough review.
Data Privacy
The privacy of data shared with generative AI tools has been a topic of concern since these tools entered the market. Part of the reason many generative AI models improve so rapidly is their use of user feedback and inputs to refine responses and expand knowledge.
Imagine an HR professional submitting sensitive career data of job candidates to an AI chatbot and asking it to categorize them based on experience. If the chatbot uses this data to improve its accuracy, some security experts warn that this sensitive information might be exposed to other users who know the right questions to ask.
Although such cases are extreme, they highlight some unresolved issues surrounding generative AI and data privacy. While many AI tools explicitly state that user data isn’t shared with others, this isn’t a universal policy.
Your company’s policy should make data privacy a core principle, specifying, for instance, that no individual or business-related data should be submitted to generative AI tools via prompts.
Security and Reputation
Beyond intellectual property and privacy concerns, your policy should also address general security risks, compliance issues, and potential reputational damage arising from accidental breaches, data leaks, or biased content.
Why Create a Generative AI Policy?
Your company should have a generative AI policy for the same reasons it has other policies—to ensure compliance, guide employees in their work, and simplify day-to-day operations.
Generative AI tools, capable of creating text, images, audio, and even video from simple instructions, have attractive applications across your business. With the variety of tools available and their rapid growth, your company risks falling behind competitors if it doesn’t start exploring how to leverage these technologies.
However, for employees to confidently experiment with these tools and for the company to avoid well-known risks, a policy is essential. Such a policy demonstrates that the company has thoughtfully considered the implications of generative AI and provides employees with clear guidelines, enabling them to safely reap the benefits of available generative AI tools.
Risks Generative AI May Pose
Unfairness, Bias, and Unethical Behavior
Generative AI tools are only as objective as the data they’re trained on. If the datasets used by a tool to create content are biased, the tool’s outputs will reflect that bias. For instance, consider a generative AI tool designed to create job descriptions based on minimal details. If the tool was trained on thousands of job descriptions containing non-inclusive language, it will produce similarly non-inclusive results.
Leading AI developers are working to reduce bias in their tools, but completely eradicating it remains a significant challenge. Your generative AI policy should acknowledge this limitation and specify measures to address it—for example, requiring all AI-generated content to be carefully reviewed by a human for bias and fairness before publication or distribution.
This article explores key risks generative AI presents, how a policy can help mitigate them, and what creating such a policy might look like for your company. While it’s not intended as legal advice, it’s based on our own experiences with generative AI tools and aims to provide a foundation for further research.
