EAST LANSING – Economist Patrick Anderson in May presented findings at an International conference that demonstrates the potential for substantial financial damages for business worldwide from the mismanagement and misuse of Artificial Intelligence and machine learning data.
The findings, which describes four main categories of damages resulting from improper implementation of AI principles, suggests a clear lack of awareness of and preparation for potential harm. These failures in management pose both organizational risks and risks to citizens’ privacy everywhere.
Anderson attended NAFE International in Evian-les-Bains, France, on May 25, the one-year anniversary of the European Union General Data Protection Regulation, the most important change in data privacy regulation in 20 years. Anderson’s company, Supported Intelligence, has given him an inside look at AI and machine learning code.
“In at least some cases,” Anderson said, “AI is misused, misapplied, conducted without proper supervision, implemented using data for which proper permission was never obtained, presented in a way that conceals its actual performance, or a combination of these.”
To mitigate these risks, Anderson proposes the development of a methodology to evaluate AI applications and data. His proposal includes steps for conducting “AI audits,” with the goal of preventing damages from ignorance and misuse that dwarf those caused by the deliberate distribution of malware.
“We had an international cast there from Italy, Germany, and the United States,” Anderson said. “It was received well. People there had their eyes opened to the serious degree of risk businesses are facing that they do not have a handle on. There are different categories of damages and how business could be on the hook for millions or hundreds of millions of dollars from machine learning usage. Customers may be running algorithms that they don’t understand.”
He said data breeches and malware are only two of five categories of damages that can emerge from misuse of these applications.
One of Anderson’s recommendations is business managers need to get a handle on AI potential damages and fast. Any company doing business with anyone in Europe that violates GDPR could be liable for a fine equal to 4 percent of worldwide revenues.
Another of his recommendations is that companies undertake an AI algorithm and data audit by an independent professional to assess whether doing the code is doing what the manager want it to do and whether it exposes customer data.
“Our recommendation is stay concerned about data breeches, be vigilant about malware, but start thinking hard about all the algorithms your company runs or are retained by your company,” Anderson said. “There are the risks that managers do not fully comprehend.”
For more information about Supported Intelligence, click on http://www.supportedintelligence.com/