“What You Need to Know about AI: Data Privacy Regulations and IP Protections,” a recent webinar hosted by the Center for Technology Commercialization (CTC), provided participants useful and timely information to protect their organizations when using artificial intelligence (AI).
Mike O’Malley of SenecaGlobal presented details of the American Data Protection & Privacy ACT (ADPPA) that was introduced in 2022 and is currently in negotiation with Congress. He explained that the act could require organizations to provide audits of their algorithms, clear explanations of data collected and usage parameters and compliance reports to the Federal Trade Commission. Very basically, groups would need to prove how safe and effective their algorithms are.
O’Malley stressed the importance of preparing now for upcoming regulation, especially since AI is a trending topic that is likely going to become even more widely used. He offered specific steps a team designing operational processes should take:
- Identify where AI is used in your organization
- Evaluate algorithms during the design phase (looking closely for bias or unintended consequences)
- Consider the data required to train the algorithm (checking for completeness and acknowledging how previous history could be different than current or future realities)
- Audit algorithms on a planned schedule to make sure they are producing intended results
- Produce reports showing outcomes and mitigation steps if needed
O’Malley provided lots of examples to make the technical information easy to understand.
Example of incomplete data: If a business only included photos of Labrador dogs in data given to their “dog” algorithm, the AI would conclude in error that no other breeds are actually dogs.
Example of unintended bias: If a predominantly male college only included resumes of students who historically had succeeded at their college in an “admissions” algorithm, the AI might not include any resumes that mentioned “women’s studies” or a traditionally female college.
O’Malley also helped audiences understand how intellectual property should be protected. He cited pros and cons to U.S. patent and trade secret protection options. Patents give the owner a virtual monopoly on the invention or intellectual property for 20 years, but trade secrets are not limited by time. Patents require an owner to disclose their “secret recipe,” but trade secrets do not. However, someone could “figure out” the trade secret by reverse engineering and then it is no longer secret or protected. O’Malley named ideas to help keep trade secrets:
- Use nondisclosure agreements for those who work with the secret
- Limit access of the trade secret
- Encrypt files
- Provide cybersecurity and other security measures
- Mark documents as confidential
Example of trade secret: The formula for Coca-Cola has been a trade secret for over a hundred years, much longer than 20 years, because they did not have to provide the recipe on a U.S. Patent, and so far no one has been able to exactly replicate the formula.
If you’re developing AI in Wisconsin or looking to implement AI in your business, reach out to CTC for assistance and to get connected with SenecaGlobal.