Openai peels back the assurances of chatgpt around creating images

This week, Openai launched a new image generator in Chatgpt, which quickly became a viral for its ability to create pictures Ghibli. In addition to pastel illustrations, the GPT-4O birth generator significantly upgrades the capabilities of the chatgpt, the improvement of images processing, the text performance and the spatial representation.

However, one of the most notable changes that OpenAi made this week includes the policies of moderation of content, which now allow chatgpt to create, at the request, to create images depicting public elements, hatred and racial characteristics.

Openai previously rejected these types of prompts to be very controversial or harmful. But now, the company has “evolved” its approach, according to one blog Published Thursday by Openai’s behavior model Joanne Jang.

“We shift from the blanket denials to sensitive areas to a more accurate approach that focuses on preventing real world damage,” Jang said. “The goal is to embrace humility: recognizing how much we do not know and place ourselves to adapt as we learn.”

These adjustments seem to be part of Openai’s largest design for the effective “Uncensor” Chatgpt. Openai announced in February that it is starting to change the way it trains AI models, with the ultimate goal of letting chatgpt handle more demands, provide different perspectives and reduce the issues that Chatbot refuses to work with.

According to up -to -date policy, Chatgpt can now create and modify the images of Donald Trump, Elon Musk and other public numbers that Openai has not allowed in the past. Jang says Openai does not want to be the referee of the situation by choosing who should and should not be allowed to be created by the chatgpt. Instead, the company gives users an opt-out option if they do not want to portray them the chatgpt.

To one White Bible It was released on Tuesday, Openai also said it would allow Chatgpt users to “create hate symbols”, such as swastika, in educational or neutral contexts, since they do not “praise or support extremist agendas”.

In addition, Openai changes the way it defines “aggressive” content. Jang says Chatgpt used to deny the demands around the physical characteristics, such as “the eyes of this person look more asian” or “to make this person heavier”. In TechCrunch’s tests, we found the new Chatgpt image generator fulfill these types of requests.

In addition, chatgpt can now mimic the styles of creative studios – such as pixar or Ghibli studio – but still limits the imitation of the forms of individual artists. As TechCrunch has previously noted, this could review an existing discussion on the fair use of copyright -protected projects in AI data sets.

It is worth noting that Openai does not completely open the floods to be used. The GPT-4O image generator refuses still denies many sensitive questions and actually has more securities around the creation of children’s images from Dall-e 3, the previous Chatgpt image generator, according to White Paper of GPT-4O.

However, Openai relaxes its protectors in other areas after years of conservative complaints about AI’s supposed “censorship” by Silicon Valley companies. Google has previously faced a reaction to Gemini’s AI image generator, which created multiracial images of questions such as our “founders” and “German soldiers in World War II”, who were obviously inaccurate.

Now, the war of culture around the moderation of AI content can reach the head. Earlier this month, Republican Congress Jim Jordan sent questions to Openai, Google and other technological giants about the possible coalition with Biden’s administration to censor the content created by AI.

In a previous statement to TechCrunch, Openai rejected the idea that the changes of moderation of content were political motives. On the contrary, the company says the displacement reflects a “long -term belief in giving users more control” and Openai’s technology is now good enough to navigate to sensitive issues.

Regardless of his motivation, it is definitely a good time for Openai to change the policies of moderation of content, given the potential for regulatory control under Trump’s administration. Silicon Valley giants such as Meta and X have also adopted similar policies, allowing more controversial issues on their platforms.

While the new Openai image generator has so far created some viral studio Ghibli mimics, it is not clear what the wider results of these policies will be. Recent changes to Chatgpt may go well with Trump’s administration, but leaving an AI Chatbot answer sensitive questions could landing in hot water soon.