Banner artwork by Moor Studio / Shutterstock.com
Though we’re seeing the global use of AI rapidly expand, most enterprises have not taken precautions to ensure their AI use is trustworthy and responsible. The 2022 IBM Global AI Adoption Index found that 77 percent of companies are using or exploring AI in their businesses. Yet, most have not taken steps to reduce bias (74 percent), track performance variations and model drift (68 percent), and ensure they can explain AI-powered decisions (61percent) — all measures that help ensure AI use is safe, responsible, and ethical.
This lack of transparency and accountability around AI use can lead to public distrust, as people question the motives of large corporations. If a company wants to be viewed as reputable and ethical, its leaders must raise visibility and accountability regarding AI. Because lawyers must maintain integrity in their practices, advocating for increased transparency around AI applications is a natural extension of the in-house legal team’s role.
AI audits and public disclosures offer valuable ways to bridge the public trust gap and proactively demonstrate commitment to responsible AI practices. Adopting the five best practices below can help you set up your AI audit program for success.
1. Collaborate with IT, AI, and data specialists
An AI audit evaluates an organization's AI systems, algorithms, and data practices to identify potential biases, privacy concerns, and other ethical issues associated with AI deployments — not traditional law school fodder.
Explainability is a core tenant of responsible AI use. One must be technologically savvy to understand and explain data collection methods, algorithmic decision-making processes, potential biases, and measures that ensure fairness and accountability.
If you’re like most lawyers, this will mean partnering with company IT, AI, and data specialists who can help you understand the inner workings of AI systems and provide core insights into their intricacies. Don’t hesitate to consult external sources with specialized knowledge in AI auditing, data privacy, and security to understand more advanced AI systems and complex use cases. Your AI audit program must be comprehensive for it to be effective.
Explainability is a core tenant of responsible AI use. One must be technologically savvy to understand and explain data collection methods, algorithmic decision-making processes, potential biases, and measures that ensure fairness and accountability.
2. Develop a robust AI audit methodology
While much of the law is theoretical and precedence, lawyers must don their scientist hats to help companies adopt robust and effective AI audit methodologies. In collaboration with tech specialists, make sure to set clear objectives and realistic timeframes. Define key performance indicators and milestones. Start by determining the specific aspects of AI systems to scrutinize, including:
- Which AI systems and tools to audit.
- What data to review.
- What specific risks to evaluate.
Your overall goals will center on identifying and managing legal and ethical risks and ensuring regulatory compliance. Regular audits allow you to proactively address any shortcomings and mitigate risks before they escalate.
3. Regularly update audits to proactively manage risk
An effective AI audit is an ongoing process, not a one-off experience. Conduct periodic AI audits to identify potential ethical or legal issues before they escalate into more complex problems. As AI systems and use cases evolve, new risks will emerge. Update audit methodologies regularly to ensure audits remain effective.
4. Enhance transparency through clear disclosures.
Consider proactively developing detailed disclosures that outline your organization's implementation of AI technologies, the potential impact on decision-making processes, and the methods used to minimize ethical or legal concerns and meet regulations.
Clear disclosures may look similar to ingredient lists and visually enhanced product labels. Make them short, simple, and easy to understand, especially for consumer-facing products and services. Publishing a summary of your AI audit results is another way to demonstrate commitment to fairness and accountability.
5. Foster an AI-literate corporate culture
AI literacy is crucial, particularly among large workforces in which rumors — and unintentional errors — easily run rampant. Educate executives and employees about AI’s potential impact on business. Help them understand how the use of AI affects their roles, specifically. Discuss ethical considerations they may encounter and the risks that can occur. Set clear policies around AI use so everyone knows what to expect and what is expected of them.
Proactively embracing these activities will help your legal teams manage the growing pressure to comply with upcoming AI governance requirements. Independent AI audits for the automated employment decision tools used in hiring and promotional decisions are at the heart of NYC 144, a groundbreaking New York law that aims to ensure AI use remains fair and unbiased. And the EU’s AI Act enforcement measures range from non-binding impact assessments to heavily audited compliance requirements.
Audit for AI trust
Conducting effective AI audits through collaboration, proactive risk management, transparency, and fostering an AI-literate culture are essential best practices to ensure compliance and maximize the benefits of this powerful technology.
