Editor: Run Layan
The FTC suddenly launched an investigation into OpenAI, and Altman tweeted 3 consecutive tweets to clarify. The FTC-OpenAI drama is slowly unfolding.
Heavy! The FTC's investigation is coming!
The survey target is none other than OpenAI, which is in the limelight.
A 20-page letter of inquiry was given Sam Altman a straight head.
OpenAI is back on the table
OpenAI was just sued last week by U.S. law firms and writers for $3 billion in damages.
Perhaps before OpenAI could choose a lawyer to help him in the case, it was approached by the Federal Trade Commission (FTC).
The FTC gave OpenAI a list of 49 big questions and more than 200 small questions, and asked OpenAI to answer and state all the questions on the list in detail.
Seeing this large series of problems, Sam Altman, who has always been known for being mature and calm, and not being alarmed, was also a little numb, and sent 3 consecutive pushes to OpenAI to suppress the shock.
The 49 questions on this list, in addition to requiring OpenAI to explain the basic company structure, are very professional and detailed for model training, privacy and prompt word risk, API plugins, personal information and other aspects.
FTC！ Be proactive
To put it bluntly, the FTC's main investigation is that OpenAI violated consumer protection laws.
The published 20-page Civil Investigation Request (CID) is not professional at first glance.
People familiar with the matter said the 20-page CID was genuine. But the FTC did not comment.
According to the CID, the FTC's investigation will focus on whether OpenAI has unfair or fraudulent privacy data security practices, and whether there is unfair or fraudulent practices that result in the risk of harm to users (including reputational damage), and whether there is a violation of Section 5 of the Federal Trade Commission Act.
For a long time, legislators have been worrying about generative AI, whether a new law is needed, how to enact, how to protect intellectual property rights, user data, is a "headache".
In this investigation, the FTC said domineeringly that we have the legal authority to pursue the harm that AI may cause.
When you think of the FTC, you have to mention Chairman Lina Khan.
As a strong woman, she is directly a shot, taking active supervision of emerging technologies rather than reacting to them.
In summary, CID requires OpenAI to list all third parties that have access to its models, and asks them to explain how they retain and dispose of user data, and where the data for their own model training comes from.
CID also asked OpenAI to explain how they assess the risks of LLM and how they detect and deal with misleading and discriminatory content.
You know, in March this year, OpenAI made an omission, when the mistake was that some users could see the titles of other users' chat history, and about 1.2% of ChatGPT Plus users' payment information was exposed.
To this end, on March 24, OpenAI published an article on its official website explaining what happened on the 20th.
The survey asked OpenAI to give a second explanation of the matter.
In fact, until now, OpenAI CEO Altman is still a sweet potato in Washington.
Lawmakers felt that Altman was fairly comfortable with legislation to regulate his own products.
However, other AI experts caution that Altman may also have his own considerations for saying this, or to listen to the different views of all parties.
Content of the inquiry letter
In this CID, the FTC lists 49 questions, covering the following aspects:
1. Basic information and background of the company
2. Relevant Notices and Representations
3. Model development and training
4. Risk assessment and response
5. Privacy and Prompt Risks and Practices
6. API Integration and Plugins
Under each category, the CID enumerates a number of issues in detail from various aspects. The more important ones are as follows:
- OpenAI must disclose its full legal name, address, phone number, and details of OpenAI's structure, subsidiaries, partnerships, etc.
- During the applicable period, OpenAI must disclose the number of employees, organizational charts, executive and director details, shareholder details, etc. annually.
- OpenAI must disclose the details of all agreements entered into with third parties for the development of LLMs during the applicable period.
- OpenAI must disclose its policies and procedures for training, testing, auditing, and monitoring LLMs.
- OpenAI must disclose the data source, data type, data retention, and details of the person/department responsible for the data used to train the LLM.
- OpenAI must disclose its technical and organizational security measures to reduce the risk that the LLM output contains illegal, unsafe, biased, or misleading content.
- OpenAI must disclose details of user complaints that have been harmed and the responses to them.
- OpenAI must disclose policies and procedures related to the privacy, security, and ethics of AI systems.
- OpenAI must disclose details of internal communication, evaluation, and research related to the safety, bias, and factual correctness of its LLM.
- OpenAI must disclose details of how it monitors compliance with policies related to auditing AI systems.
- OpenAI must disclose details of instructions, guidelines, and training materials provided to LLM users.
- OpenAI must disclose any external research/testing related to its LLM and outline the results.
- OpenAI must disclose details of any modifications it makes to LLM in response to security issues.
Looking at this series of detailed and comprehensive questions, if OpenAI really has to make a detailed statement of the content inside, in case the content is "accidentally" leaked, OpenAI may really be completely OPEN.
The house leak coincided with overnight rain
OpenAI's recent troubles really don't stop there.
On June 28, the Northern California law firm Clarkson filed a class action lawsuit against OpenAI on behalf of 16 plaintiffs.
It alleges that OpenAI grossed data from the Internet while training large models for serious infringement of the copyrights and privacy of 16 plaintiffs, and seeks $3 billion in damages.
The central claim of Clarkson's lawsuit is that OpenAI's entire business model is based on theft.
A few days later, four authors, Mona Awad, Paul Trembla, Sarah Silverman, Christopher Golden, and Richard Kadrey, also sued OpenAI in different U.S. district courts, accusing them of violating copyright law by using their books to train ChatGPT without consent.
Seeing this, some readers may want to ask, didn't you write about Sam Altman vs. Capitol Hill before and won regulatory support, how did so many people find OpenAI stubble at once?
The main problem here is that there are three groups of people in charge of affairs in American society: the old men who legislate in Congress, the judges who hear cases in the courts, and the executive branch of the government.
So Sam Altman won the support of the first group of people, the old men of Congress who legislated, but he still hadn't managed the other two sets of relationships.
The matter of the past two days belongs to these two groups of people who have not been solved.
And the FTC who came to check the water meter today should be the most difficult of all people to deal with.
Tech company rival – FTC
FTC Chinese full name is Federal Trade Commission, which functions similar to our nation's Department of Commerce.
Readers who have paid a little attention to the news of the US technology circle in recent times should be familiar with this institution, in addition to the trouble of OpenAI, its achievements also include: investigating and punishing some cryptocurrency exchanges such as FTX, and fighting an antitrust lawsuit with Microsoft.
Yesterday, the FTC's motion to temporarily suspend Microsoft's acquisition of Blizzard was just rejected by the court, and today I went to check OpenAI's water meter.
I have to sigh that the efficiency of the FTC, a government agency, is comparable to that of a startup.
Looking at the history of the FTC's previous confrontations with technology companies, Facebook, Google, Microsoft, and Amazon have all been punished by it for various problems, and it can be said that they are the old sufferers of technology companies.
Its current chairman, Ms. Lina Khan, is quite a legend.
She was 32 years old when Biden appointed as FTC chairman on June 15, 2021.
How to say this experience, in China, it is equivalent to a 32-year-old lady becoming the minister of commerce, which belongs to the oversized hostess who the author of the cool article does not dare to stand.
In addition to being entrusted with important responsibilities at a young age, she also debuted with the aura of "anti-big tech companies".
In 2017, Lena Khan, who was still studying, published an article in the Yale Law Journal entitled "Amazon's Antitrust Paradox", which caused great repercussions in legal and political circles.
She served as counsel to the House Judiciary Committee's Antitrust Panel, where she helped write a 449-page report accusing Amazon, Apple, Facebook and Google of abusing market dominance.
So less than a month after she was appointed chairman of the FTC, Google and Facebook petitioned the FTC to recuse herself from antitrust lawsuits against tech companies.
These two rich and rival countries, calling for wind and rain in Silicon Valley, are sincerely afraid of the new leader, worried that they really can't beat it, so they can only ask the system to ban her first.
It's conceivable that as long as she's still in office, big tech companies won't have a good time.
And OpenAI because of the status of "Silicon Valley's popular fried chicken" as a start-up company was targeted by the FTC, which may be regarded as an alternative affirmation of their status.
Netizen reaction
After Altman posted three tweets in the heart of the stable army, netizens below had different opinions.
However, after roughly sweeping over, it was basically all criticism of Altman. Is it really "the world has suffered OpenAI for a long time"?
For example, the following netizen, this is all euphemistic.
"I know you want users to be happy with your product, but your technology is yours, and how you use it is up to you. I think you have to be honest and admit that you personally have much more exposure to models than users."
"Your honesty is very important."
This netizen is more blunt, and it is estimated that he can directly silence Altman.
"I use GPT every day, why don't I think you protect the rights and interests of users?" Without our consent and collecting information, how dare you say that you are defending your rights and interests?"
The following netizen directly opens Yin and Yang:
"How fresh! A company, saying uh-huh, we absolutely abide by the law, in fact, all user information has been leaked, believe you a ghost. Your OpenAI is awesome."
"If Altman had said that the rhetoric about AI security was just for his own sake, I wouldn't have listened."
However, some netizens on OpenAI said that they were very uneasy about the regulators, worried that they would not be able to manage this matter well.
The fear is that biased regulators can't really regulate effectively.
This netizen is even more lively and does not think that it is too big, hoping that after OpenAI hands over the key data to the FTC, it will be leaked in some form, and finally complete the "open source" of OpenAI.
This last comment must have spoken the hearts of many people. Although that CID looks professional, the FTC may not have enough professional personnel to follow up.
Even evaluating OpenAI's answer can be laborious.
This is instructive – professional legislation "can and only" be done by professionals.
And it seems that this netizen does not have much confidence in the FTC.
Resources:
https://www.nbcnews.com/tech/tech-news/ftc-investigating-chatgpt-maker-openai-possible-consumer-harm-rcna94067