FTC Starts Inquiry of OpenAI Data Security Methods

In May, Sam Altman, the OpenAI co-founder, made the rounds in Washington DC encouraging lawmakers in Congress to regulate generative AI technology before the chatbots get too smart and start regulating humans, or worse.

Be careful what you wish for, Sam.

The Federal Trade Commission has opened an investigation into OpenAI focused on whether the startup has harmed consumers through its dissemination of false information as well as private personal data by ChatGPT.

Federal regulators sent the company a 20-page letter last week notifying OpenAI they’re examining if it “engaged in unfair or deceptive privacy or data security practices or engaged in unfair or deceptive practices relating to risks of harm to consumers,” according to a report in the New York Times.

At a hearing of the House Judiciary Committee on Thursday, Lina Khan, the FTC chair, said “ChatGPT and some of these other services are being fed a huge trove of data. There are no checks on what type of data is being inserted into these companies.”

Khan added that there have been reports of people’s “sensitive information” showing up on the GPT bots.

Since its release in November, ChatGPT has taken the world by storm: the GPT bots (GPT-4 came out in March) now are accessed by an estimated 200M people worldwide. While GPT-4 was a big leap forward from ChatGPT, users report that the bots have been known to blend fact with fiction and make stuff up—the scientists call it “hallucinating.”

Altman responded to the FTC’s probe with a series of tweets: “it is very disappointing to see the FTC’s request start with a leak and does not help build trust. that said, its super important that out technology is safe and pro-consumer, and we are confident we follow the law. of course we will work with the FTC,” he said.

Altman also said “we protect user privacy and design our systems to learn about the world, not private individuals. we’re transparent about the limitations of our technology, especially when we fall short.”

Syntax, grammar and no caps aside, it’s curious to hear Altman declare he’s confident OpenAI “follows the law” after he just met with dozens of lawmakers telling them they urgently needed to write the first laws governing generative AI.

Altman has been quite—there’s no other way to say it—open about OpenAI’s method of training its chatbots.

The Large Language Model for the GPT platform was created by feeding the digital neural network large chunks of the Internet, including the biases and misinformation that are available in abundance on the dark side of the Web.

OpenAI, which now has more than $10B in backing from Microsoft, has admitted it’s conducting a global Beta test of the bots, which will continue to get smarter as they digest the queries of hundreds of millions of end-users.

In March, the Italian government banned ChatGPT, saying OpenAI unlawfully collected personal data from users and did not have an age-verification system in place to prevent minors from being exposed to illicit material. OpenAI restored access to the system the next month, saying it had made changes Italy had requested.

The Center for AI and Digital Policy, an advocacy group, has called upon the FTC to block OpenAI from releasing new commercial versions of GPT, citing concerns involving bias, disinformation and security.

A class-action lawsuit filed at the end of June in federal court in San Francisco accuses OpenAI of stealing “vast amounts” of personal information, intellectual property and copyrighted content to train its chatbots. The lawsuit says OpenAI violated privacy laws by “secretly” scraping 300 billion words from the Internet, including “books, articles, websites and posts.”

The lawsuit, which also names Microsoft as a defendant, accuses Open AI of risking “civilizational collapse.” It seeks $3B in damages.

“Despite established protocols for the purchase and use of personal information, Defendants took a different approach: Theft,” the lawsuit, filed by the Clarkson Law Firm, alleges. The suit also cites claims of invasion of privacy, larceny, unjust enrichment and violations of the Electronic Communications Privacy Act.