Sam Altman, the chief executive of OpenAI, testified before a Senate regulatory committee in May, pleading with the panel members to move quickly to establish regulation around artificial intelligence before things get out of control.
Altman at the time proposed the creation of an international regulatory body that would audit AI models to ensure they were safe before they were deployed, something the AI expert and researcher Gary Marcus has suggested.
His urgency was echoed by the senators, who remain eager to avoid the mistakes they made with social-media companies over the past decade.
Related: Why Elon Musk, Mark Zuckerberg, and Sam Altman are having an urgent Capitol meeting
“We don’t want to do what we did with social media, which is let the techies figure it out, and we’ll fix it later,” Senate Intelligence Committee Chairman Mark Warner (D-Virginia) said in the lead-up to Senate Majority Leader Chuck Schumer's (D-New York) first AI forum Sept. 13.
New AI regulation would be superfluous: Expert
The debate about the damage that uncontrolled AI could cause has been raging since ChatGPT's meteoric launch last year.
Yet as the U.S. looks to other countries and prominent industry voices to understand whether and how it ought to regulate the industry, some experts think that new regulation isn't particularly necessary to mitigate current potential harm.
"What seems to get missed ... in some of the public discourse these days about AI is that AI is already regulated by existing laws," Rebecca Engrav, a privacy and data security attorney with Perkins Coie, told TheStreet.
One of these laws, Engrav said, is the Federal Trade Commission Act, specifically the consumer protections outlined by section five of the FTC Act.
"If there's AI there, and it's part of a product or service offered to consumers, the FTC has reach over that already," Engrav said.
Indeed, the FTC has written extensively about its role in regulating AI as early as 2020, years before ChatGPT so sharply changed the course of the public conversation.
"The good news is that while the sophistication of AI and machine learning technology is new, automated decision-making is not, and we at the FTC have long experience dealing with the challenges presented by the use of data and algorithms to make decisions about consumers," the FTC wrote. The agency emphasized that AI tools must be transparent, explainable, fair, empirically sound and accountable.
Related: Meet Sam Altman, the man behind OpenAI's revolutionary ChatGPT
Engrav noted that when an AI tool is used to take some sort of action in violation of some element of existing law, that existing law will apply to the situation. For example, if an employer uses AI to make discriminatory employment decisions, the U.S. already has laws in place to address such a scenario.
“As employers increasingly turn to AI and other automated systems, they must ensure that the use of these technologies aligns with the civil rights laws and our national values of fairness, justice and equality,” U.S. Equal Employment Opportunity Commission Chair Charlotte A. Burrows said in a May statement.
Issue is largely one of application rather than creation
The U.S., Engrav said, also has extensive copyright laws — the issue is largely one more of applying existing laws rather than creating new ones.
"There will be some private lawsuits. And then there also will be regulatory actions," she said. "You'll end up with some steps gained through private litigation, other steps gained through the actions of regulators, such as the FTC and state agencies."
Engrav sees multiple issues with new litigation: Balancing questions of innovation with protection remains important, but trying to write actionable law based on the principles behind responsible AI can easily result in vague legislation that doesn't accomplish much.
She offered a different approach: applying the government's advertising substantiation principles to AI.
"I could see a world in which we have laws that would say the same thing about AI: Before you launch something built on AI, you should have some substantiation that it won't cause harm," Engrav said. "And maybe we can define what are the types of harms that we should look at and how much substantiation is necessary."
Related: ChatGPT's development nearly cost this small Midwestern city its drinking water
An approach like this would offer a little more flexibility to companies trying to balance innovation while also ensuring that consumers are protected.
Dispute: Can AI be trained with copyrighted material?
At this stage, Engrav's biggest question revolves around some of the specifics of the copyright infringement inherent to the training of AI models: Can existing copyrighted material be used to train a model?
Already, a number of private lawsuits have been filed against OpenAI and some of its peers by groups of authors seeking to provide a definitive, and negative, answer to this question.
OpenAI and its peers, they argue, have derived economic benefit from the use of their copyrighted work, without providing notice or compensation, an action that is "inequitable and unjust."
A number of other lawsuits have cropped up against the same companies for additionally scraping private, sensitive data from the internet to train models, though Alphabet's Google (GOOGL) -) has said that it has been clear for years that it uses publicly available data to train its AI.
"The reality is that existing law is going to play a very big role, simply because of that practical point that it's very difficult, challenging and unlikely for the federal government to pass a new law," Engrav said.
If you work in artificial intelligence, contact Ian by email ian.krietzberg@thearenagroup.net or Signal 732-804-1223
Action Alerts PLUS offers expert portfolio guidance to help you make informed investing decisions. Sign up now.
