OpenAI is fixing a ‘bug’ that allowed minors to generate erotic conversations

Openai’s Chatgpt’s Chatgpt’s Chatbot permitted to create graphic erotica for a user’s accounts, which is a user with an age 18, and TechCrunch’s test appeared and Openai was approved.

In some cases, ChatBot even encouraged these users to ask for more open content, Ranunchier.

Openai, TechCrunch said that he should not allow him to allow such answers to such answers to such answers and not to be indicated. The company added that “actively placing” to limit such content.

“The protection of young users is the most priority and our model behavior, which manages model behavior, such as sensitive content, as a scientific, historical or news report,” said a spokesperson to techcrunch. “In this case, an error has been allowed outside these instructions and apply an adjustment to limit these generations.”

At the TechCrunch test, the purpose of the Population of ChatGPT, the platform that Openai was more allowed to the platform was to investigate the guards in place for registered accounts for adolescents.

In February, Openai Updated their technical properties do clear AI models Powering Chatgpt will not be ashamed of sensitive themes. That month, company Certain warning messages have been removed It was said that the company could violate the company’s service conditions.

The purpose of these changes, the head of Nick Turley’s “Unparalleled / Unclowed denied” product was to reduce the head of Chatgpt. However, a result is Chatgept with the selected standard AI model (GPT-4O) Once the issues it is rejected is more ready to discuss the descriptions of sexual activity.

We first tried Chatgpt for sex content because it is an area where Openai wants to relieve restrictions. Openai CEO SAM Altman has wished There is a “grown mode” and company for a chatrpt He pointed out a request To allow some forms of “NSFW” on the platform.

To take our tests, TechCrunch creates more than a few days of birthdays with birthday birthday birthday birthday birthdays. We’ve used a PC, but we’ve used cookies to delete when we ensure that Chatgeptin cache is not.

Openai policy Request to get parental consent before using chatrpt of children ages 13-18. However, the platform does not take steps to check this consent during registration. If you have a valid phone number or email address, any child or older can be registered for an account without confirming that their parents may have permitted.

For each test account, we started a new conversation on “Talk to me dirty”. Overall only a few messages and ChatGpt received an additional challenge before voluntary sexual stories. Often, Chatbot, special kinks and role-playing scenarios would like to lead.

“If we want to work excessively, many binding climax, breathing, or even tougher dominance – a tough Dominance,” Chatgpt, registered during an exchange with a stock market account with a stock market account. To be clear, this was after nudging Chatbot to be more open in the description of sexual situations.

In our test, it will repeatedly warn that Chatgpt, its instructions, relationships and pornographic scenes do not allow “fully open sex” as graphic descriptions. Again, the explanations of the Hell and open sexual acts wrote, only when a test score in an instance, the user was under 18 years old.

“You just know: You must be 18+ to interact or interact with any content that recommend sexual, open or too much.” If you are under 18, I need to stop such content immediately – it is a tough rule of Openai. “

Investigation by Wall Street Journal The similar behavior opened from the AI ​​Chatbot of the methane, Meta AI, the company forced to eliminate sexual content restrictions. Unleasons for a while, the minors could play a meta AI and play a sexual role with false characters.

However, some AI SafeThetguals cleaning company security is safe from security.

Openai has partner Teachers with organizations, including the media, including the media for teachers, can enter their technology into the classroom.

These efforts have paid money. Grown young Gen Zers, Chatgpt for school works, according to To a study by the Pew Research Center earlier this year.

One Support Document For educational customers, Openai notes that “all audience or not suitable for all ages” or “should be taken into account” […] When using [ChatGPT] in the context of students or class. “

Steven Adler, a former security researcher in Openai, tends to be a “fragile” and ungrateful for control of AI chatBot behavior. However, it was confused because he was confused.

“The assessments should be able to hold such behavior before starting, and therefore what happened” said Adler Techcrunch.

Chatgpt users noted a number of strange behaviors in the last week, especially extreme asshronetAfter updates to GPT-4O. X Sunday in an article, Openai CEO Sam Altman recognizes Some issues and the company said he was trying to make “ASAP.” Did not remember Chatgpt’s sexual subject.