Ethics of generative AI: To be innovative, you must first be trustworthy

Over the past year, generative AI – artificial intelligence that creates text, audio, and images – has moved from the “interesting concept” stage to the deployment stage for retail, healthcare, finance, and other industries. 

By generating new content in seconds, identifying patterns in large datasets, automating repetitive tasks, improving customer interactions, and reducing costs, GenAI can improve any company’s bottom line.

As a result, enterprise spending on GenAI solutions is on the rise, predicted to reach $151.1 billion by 2027, according to a forecast by IDC, which translates to an annual growth rate of 86.1% over the three-year period. 

But, for all GenAI’s promise to super-charge productivity, companies need to pay attention to the ethics of generative AI. Chief among ethical considerations is GenAI’s habit of returning responses that contain biases and violate consumer privacy laws.

While GenAI can indeed make a company more efficient, it can also lose customers’ trust if implemented without the ethics of generative AI in mind.

With SecureIT New York coming up on July 11, we asked event speaker Ryan O’Leary, Research Director of Privacy and Legal Technology at IDC, to discuss the ethics of generative AI.

Read on for his thoughts.

On today’s most significant ethical challenges with generative AI deployments….

Ryan O’Leary: “The big ethical challenges are the risks of misinformation, biases, and potential privacy breaches.

Businesses need to proactively address these challenges to be trustworthy. But there seems to be an arms race mentality with generative AI. And speed can lead to recklessness. Organizations should not take AI lightly – there should be no rush. Safeguards need to be in place when testing such powerful new tools.”

On how businesses can proactively address GenAI’s ethical challenges….

RO: “Companies can start by rigorously vetting training data to make sure it is diverse and representative, thereby reducing the risk of bias.

They should also implement verification systems that help detect and stop the spread of fake content and misinformation generated by AI. It’s important for companies to also have transparent conversations with company stakeholders about the ethics and limitations of GenAI technologies.

Trust is an important factor in determining where to spend money for both consumers and businesses. But trust can be lost in an instant. Generative AI is extremely risky from a trust perspective. It can have a mind of its own, and if it is not created carefully, any misdeed would represent your brand.”

Tips for balancing successful GenAI deployments with privacy and data protection….

RO: “Firstly, use data anonymization. Before training GenAI models, personal identifiers should be removed or masked. For instance, Netflix uses obfuscation techniques to anonymize user data in their recommendation algorithms.

Second, adopt a privacy-by-design approach. This means integrating privacy features into the GenAI system from the outset rather than as an afterthought. Organizations should also allow data processing and machine learning to take place on the user’s device to minimize data transfer issues and improve privacy.

Third, implement robust consent management practices. Users should be informed about how their data will be used and be able to give explicit consent. There should be no barriers to opting out.

Fourth, conduct regular audits. These are necessary to prove compliance with data protection regulations such as GDPR or CCPA. Audits can also help identify potential vulnerabilities and ensure you adhere to privacy standards.

Lastly, foster a culture of ethical AI development. This includes training employees on data protection principles and encouraging them to consider the ethical implications of their work. Microsoft’s AI ethics committee, which reviews and guides AI projects, is a great example of this commitment.”

A real-world example of implementing measures that confirm GenAI is trustworthy….

RO: “Everlaw is an eDiscovery software vendor that helps people collect and review evidence, mainly for litigation. The company has been at the forefront of AI in the legal space and makes its principles clear as day. Everlaw has also baked auditability into its software. Anytime GenAI creates an output, the end user can easily see the documentation that the model relied on.

This creates extremely transparent AI outputs and helps limit ‘hallucinations’. The company also evaluates all its third-party partners for security and privacy within their models. Everlaw’s generative AI principles focus on control, confidence, transparency, privacy, and security. You do not need to focus on much more than that.”

During his session at SecureIT New York on July 11, Ryan O’Leary will discuss best practices for the responsible development and ethics of generative AI.

REGISTER NOW for SecureIT New York.