AITechnologyAbsract_G1148091792_Feature_1920x945

04.27.21

The Federal Trade Commission releases new guidelines on Artificial Intelligence

BY Tracy Ickes

In a blog post last week, the Federal Trade Commission (FTC) makes it clear that it will consider prosecuting algorithm-based bias under the Equal Credit Opportunity Act, the Fair Credit Reporting Act, and as an unfair or deceptive practice under the FTC Act.

The FTC encourages developers to consider whether all populations are represented in data sets and to test algorithms during and after development to make sure they do not discriminate on the basis of race, gender, or other protected classes. So that biases can be identified, the FTC encourages transparency, including publishing the results of independent audits and allowing inspection of data or source code. At the same time, the FTC warns that exaggerating an algorithm’s ability to deliver fair or unbiased results is, itself, a deceptive act that may be subject to FTC enforcement action.

The FTC also cautions businesses to “tell the truth about how you use data,” citing pending FTC complaints regarding disclosure (or alleged lack thereof) to users whose photos were used in facial recognition algorithms.

And noting how broadly it may reach, the FTC cautions, “If your model causes more harm than good—that is, in Section 5 parlance, if it causes or is likely to cause substantial injury to consumers that is not reasonably avoidable by consumers and not outweighed by countervailing benefits to consumers or to competition—the FTC can challenge the use of that model as unfair.”

Lest there be any doubt that the post is a warning to AI developers and users, the post warns, “Hold yourself accountable—or be ready for the FTC to do it for you.”

The Supreme Court just ruled on Thursday that the FTC is not authorized to seek monetary relief for violations of the FTC Act, but it does retain the power to seek injunctive relief. Thus, the FTC may still seek aggressive relief such as deletion of data obtained through deceptive means, or even deletion of models or algorithms altogether. Meanwhile, legislative hearings on increasing the FTC’s authority are already underway. Thursday’s ruling from the Supreme Court does little to undermine the strong position taken by the FTC against algorithm-based bias.

author img

Author

Tracy Ickes

Associate

Posts By this author