Federal Trade Commission Begins Investigation Into ChatGPT Maker
The focus of probe into OpenAI is on the potential harm to consumers and data security concerns.
The Federal Trade Commission (FTC) is opening an investigation into the artificial intelligence firm OpenAI, the maker of ChatGPT.
The focus of the probe is on whether or not the chatbot has harmed consumers through its collection of data, its publication of false information about individuals, and its questionable data insecurity.
In a 20-page letter sent to the San Francisco company this week, the agency said it was also looking into OpenAI’s security practices. The F.T.C. asked OpenAI dozens of questions in its letter, including how the start-up trains its A.I. models and treats personal data, and said the company should provide the agency with documents and details.
The F.T.C. is examining whether OpenAI “engaged in unfair or deceptive privacy or data security practices or engaged in unfair or deceptive practices relating to risks of harm to consumers,” the letter said.
This investigation represents the first serious U.S. regulatory threat to OpenAI.
One question asks the company to “describe in detail the extent to which you have taken steps to address or mitigate risks that your large language model products could generate statements about real individuals that are false, misleading or disparaging.
The new FTC investigation under Chair Lina Khan marks a significant escalation of the federal government’s role in policing the emerging technology.
Khan, who appeared before the House Judiciary Committee on Thursday, said the agency is concerned that ChatGPT and other AI-driven apps have no checks on the data they can mine.
“We’ve heard about reports where people’s sensitive information is showing up in response to an inquiry from somebody else,” Khan said. “We’ve heard about libel, defamatory statements, flatly untrue things that are emerging. That’s the type of fraud and deception that we are concerned about.”
The inquiries are not unreasonable. OpenAI was initially presented as a non-profit organization supportive of the fair use of data for research and development.
The worry is that it may have morphed into a commercial enterprise without providing just compensation for the materials it offers.
Recently, comedian Sarah Silverman and two bestselling novelists have sued both Meta and OpenAI, asserting that both firms used the authors’ copyrighted books without permission to “train” their artificial intelligence software programs.
The proposed class-action lawsuit was filed in a San Francisco federal court on Friday by authors Richard Kadrey, known for his supernatural horror series “Sandman Slim,” and Christopher Golden, along with Silverman, who, aside from acting, published the bestselling memoir “The Bedwetter” in 2010. Each suit seeks just under $1 billion in damages, according to court filings. The authors alleged the two tech companies had “ingested” text from their books into generative AI software, known as large language models, and failed to give them credit or compensation.
The suit arrives several weeks after bestselling authors Mona Awad and Paul Tremblay also sued OpenAI for copyright infringement on similar grounds. The complaints have been filed by attorneys Joseph Saveri and Matthew Butterick, who also are behind lawsuits against controversial AI art tool Stable Diffusion on behalf of several visual artists and a proposed class action against Microsoft’s GitHub Copilot.
Representatives for Meta and OpenAI did not immediately respond to The Times’ requests for comment on Monday.
Data security is another significant issue that the FTC will be reviewing in this investigation.
The FTC also asked the company to provide records related to a security incident that the company disclosed in March when a bug in its systems allowed some users to see payment-related information, as well as some data from other users’ chat history. The FTC is probing whether the company’s data security practices violate consumer protection laws. OpenAI said in a blog post that the number of users whose data was revealed to someone else was “extremely low.”
The FTC declined to comment. OpenAI CEO Sam Altman said in a tweet Thursday evening that the company will “of course” work with the agency.
“it is very disappointing to see the FTC’s request start with a leak and does not help build trust,” he tweeted. “that said, it’s super important to us that out technology is safe and pro-consumer, and we are confident we follow the law.”
DONATE
Donations tax deductible
to the full extent allowed by law.
Comments
TY Leslie… Consumer protection may be the legal chink in their armor.. Ron Coleman has mentioned that a few times, wr to social media and censorship.
My own opinion of ChatGPT pretty much is garbage in garbage out. I have no doubt the thing was taught with a severe bias. I hope the FTC can showcase that somehow.
I expect the concern is more that it is not entirely under the control of the state bureaucrats.
Sadly though, kids in school are using if for a brain crutch. Even the Chronicle of Higher Education is noting that it is a scourge on learning.
There is a debate on whether it is detectable or not.
There is another debate on whether the chips come from China, and that’s a BIG problem.
It seems to be pretty detectable. My school had 6 seniors get busted using it to write papers for then last school year.
I hope so. There are some articles showing increasing sophistication in how CrapGPT is misused.
The chips come from Taiwan, with a few from Korea. AI right now is powered by nVidia’s tensor core GPUs which are all manufactured in either Samsung 8nm nodes or TSMC 4nm nodes.
Now, that does pose a couple of other problems, including nVidia defacto monopoly, and all of the fabs being within bombing range of China, but China cannot make those nodes.
AI….look at the first word, artificial.
Are we that stupid?
Word Salad
-noun
1) A nonsensical assemblage of words, typical of schizophrenia, Wernicke’s aphasia, and some other mental disorders.
2) Meaningless text generated by a computer.
3) The jumble of incoherent speech, as is sometimes heard in schizophrenia cases or daily by Kamala Harris, the newly appointed AI Czar.
Michele Boldrin has an entertaining counter-intuitive interview that intellectual property is harmful and unnecessary.
https://www.econtalk.org/boldrin-on-intellectual-property/
So there might be rent-seeking. Certainly here there is.
I can take your text and compute eigenvectors and eigenvalues in some high dimensional space and publish them. A bunch of numbers. I could list the most salient clusters, which would serve as sort of a gloss of your work. I could use the clusters to produce random text in your style.
Obviously ingesting isn’t the problem (see the first in the list).
The second-to-last is sort of a book review. Certainly under fair use.
The last could be parody, also fair use.
I call rent seeking.
Random letters (used for practicing high speed Morse code)
JBHRJ NKGAR YVHZA YMEIL DTXGB LSCCM ALNKC WXOEZ FCXOD XNPBX CGQZM TLGXP SAAHK EFHVL JANGP
Being random, they could have said something intelligible, false and defamatory, like Shakespeare’s thousand monkeys.
It has exactly the same mens rea as chatgpt though, namely none.
I think the argument here is something else. The creators took material that was not in the public domain and essentially photocopied it for public use. (One has to ask how they obtained a machine-readable copy of the material in the first place.)
Sure, the complaint is rent-seeking, but rent-seeking is the entire basis of copyright law.
“OpenAI CEO Sam Altman said in a tweet Thursday evening that the company will ‘of course’ work with the agency.”
Considering the extreme aggressiveness of the federal agencies in the last 12 months or so, I’d like to propose you for a medal for extreme bravery: “The Camel’s Nose Under the Tent Flap”.
Kash Patel
@Kash Elon Musk is is big tech colluding with our government to censor our elections. You’ve had months to set the record straight, n you have failed. Your cheap Titter posts and your Mickey Mouse clown droppings do not absolve you- you are as bad as FBI/DOJ n you and are making millions from the disinformation campaigns. You are a complete and total fake who cares only about $. Any time you want a real challenge, I’ll
debate you, name the place and time.
Elon = Schiff – guantlet thrown
Recently, comedian Sarah Silverman and two bestselling novelists have sued both Meta and OpenAI, asserting that both firms used the authors’ copyrighted books without permission to “train” their artificial intelligence software programs.
So … you need the author’s permission to read their books, now?
Did these authors also get permission from all the authors of the books that they read (and poached from, as is human – intentionally or not)? I wonder how many phrases in Sarah Silverman’s book can be found in books that she read? Probably something on the order of … all of them. I doubt she has actually made up any English phrases all on her own. But even if she did come up with one or two phrases, that still leaves the bulk of her writing that is nothing but copying other people’s terms and expressions.
Barns burned to the ground, the smoke cleared, let’s call the Fire dept.
The developers using chatbot, a machine, to dredge both private and public confidential data in hyper-industrialized form. They then disseminate that data via a third party. In this case the third party is an amoral machine.
In my minds eye it, albeit an impressionistic eye, I compare 19th century hydraulic gold mining technics to modern data mining. Things done “en masse.” Not scientific! and people may object to the characterization. But previous to hydraulic mining men used shovels and picks. Hydraulic nozzles automated mining.
There’s an interesting thing in this. The First World War was called the first industrialized war or termed industrial warfare.
Mechanized industrial warfare. Murder en masse.
Chatbot is a machine. A mechanized Industrial Information Warfare machine. Call it what you will; mechanized Industrial Cultural warfare. Whatever.
We were already under sustained cultural assault. That assault is now mechanized to a degree unfathomable.
Our existential big tech enemy have deployed their new machines. Those machines designed to destroy us.
This is why Musk seems to be rushing the rollout of X AI. The only way to hold off the enemy is to mechanize the defense.
Raw data to an AI is what ammunition is to a machinegun. Or so it seems. This is why Musk shut Twitter. The enemy were crawling through his database.
I still think Musk is Knight Errant in a slowly unfolding eucatastrophe.
It always comes down to claims.
I don’t believe the FTC will find anything that has OpenAI saying the results are accurate. It’s really a demonstration project.