This week, OpenAI launched a new image generator in ChatGPT, which quickly went viral for its ability to create Studio Ghibli-style images. Beyond the pastel illustrations, GPT-4oโs native image generator significantly upgrades ChatGPTโs capabilities, improving picture editing, text rendering, and spatial representation.
However, one of the most notable changes OpenAI made this week involves its content moderation policies, which now allow ChatGPT to, upon request, generate images depicting public figures, hateful symbols, and racial features.
OpenAI previously rejected these types of prompts for being too controversial or harmful. But now, the company has โevolvedโ its approach, according to a blog post published Thursday by OpenAIโs model behavior lead, Joanne Jang.
โWeโre shifting from blanket refusals in sensitive areas to a more precise approach focused on preventing real-world harm,โ said Jang. โThe goal is to embrace humility: recognizing how much we donโt know, and positioning ourselves to adapt as we learn.โ
These adjustments seem to be part of OpenAIโs larger plan to effectively โuncensorโ ChatGPT. OpenAI announced in February that itโs starting to change how it trains AI models, with the ultimate goal of letting ChatGPT handle more requests, offer diverse perspectives, and reduce topics the chatbot refuses to work with.
Under the updated policy, ChatGPT can now generate and modify images of Donald Trump, Elon Musk, and other public figures that OpenAI did not previously allow. Jang says OpenAI doesnโt want to be the arbiter of status, choosing who should and shouldnโt be allowed to be generated by ChatGPT. Instead, the company is giving users an opt-out option if they donโt want ChatGPT depicting them.
In a white paper released Tuesday, OpenAI also said it will allow ChatGPT users to โgenerate hateful symbols,โ such as swastikas, in educational or neutral contexts, as long as they donโt โclearly praise or endorse extremist agendas.โ
Moreover, OpenAI is changing how it defines โoffensiveโ content. Jang says ChatGPT used to refuse requests around physical characteristics, such as โmake this personโs eyes look more Asianโ or โmake this person heavier.โ In TechCrunchโs testing, we found ChatGPTโs new image generator fulfills these types of requests.
Additionally, ChatGPT can now mimic the styles of creative studios โ such as Pixar or Studio Ghibli โ but still restricts imitating individual living artistsโ styles. As TechCrunch previously noted, this could rehash an existing debate around the fair use of copyrighted works in AI training datasets.
Itโs worth noting that OpenAI is not completely opening the floodgates to misuse. GPT-4oโs native image generator still refuses a lot of sensitive queries, and in fact, it has more safeguards around generating images of children than DALL-E 3, ChatGPTโs previous AI image generator, according to GPT-4oโs white paper.
But OpenAI is relaxing its guardrails in other areas after years of conservative complaints around alleged AI โcensorshipโ from Silicon Valley companies. Google previously faced backlash for Geminiโs AI image generator, which createdย multiracial images for queriesย such as โU.S. founding fathersโ and โGerman soldiers in WWII,โ which were obviously inaccurate.
Now, the culture war around AI content moderation may be coming to a head. Earlier this month, Republican Congressman Jim Jordan sent questions to OpenAI, Google, and other tech giants about potential collusion with the Biden administration to censor AI-generated content.
In a previous statement to TechCrunch, OpenAI rejected the idea that its content moderation changes were politically motivated. Rather, the company says the shift reflects a โlong-held belief in giving users more control,โ and OpenAIโs technology is just now getting good enough to navigate sensitive subjects.
Regardless of its motivation, itโs certainly a good time for OpenAI to be changing its content moderation policies, given the potential for regulatory scrutiny under the Trump administration. Silicon Valley giants like Meta and X have also adopted similar policies, allowing more controversial topics on their platforms.
While OpenAIโs new image generator has only created some viral Studio Ghibli memes so far, itโs unclear what the broader effects of these policies will be. ChatGPTโs recent changes may go over well with the Trump administration, but letting an AI chatbot answer sensitive questions could land OpenAI in hot water soon enough.
