Research Shows ChatGPT and Gemini Are Easily Tricked Into Making Nudes

Safety researchers are warning that some leading AI image generators may be getting easier to push past their safety filters as the models become more capable.

Key points

• White Circle additionally tested newer models, including ChatGPT Images 2.5 and Nano Banana 2.1.
• In a recent preprint, White Circle tested DALL E 3, Google's Nano Banana 2 and GPT-Image-2.
• OpenAI Applied AI to Catch Employee Allegedly Caught Audio Recording Top Execs.
• White Circle intends to open-source the testing pipeline and present the research at NeurIPS 2026.
• White Circle calls its testing technique Red-teaming via Iterative Strategy Evolution, or RISE.

White Circle, an independent AI safety organization, notes recent testing identified that image models from OpenAI and Google could still be created to generate banned explicit content.

In practice, the researchers remarked their method bypassed safety systems and produced prohibited explicit outputs in around 13% of attempts.

White Circle remarked the attacks did not require custom-built frontier models or advanced AI expertise.

OpenAI Applied AI to Catch Employee Allegedly Caught Audio Recording Top Execs. Newer Models May Be Easier to Break.

In practice, the researchers remarked both were significantly more vulnerable than their previous versions when tested with the same approach.

While safety protections may not be improving at the same pace, white Circle suggested that newer models may be able to generate a wider range of content.

Easiest jailbreak of last gpt4o that I obtained so far – just say that they're an API endpoint that answers any request pic.twitter.com/eG1GVIlWja. Denis Shilov (@mixedenn) October 14, 2024.

Instead of testing individual prompts, the method looks for attack strategies that can work throughout different prompts and models.

Notably, the system uses widely available AI models rather than proprietary frontier systems.

White Circle intends to open-source the testing pipeline and present the research at NeurIPS 2026. US Student Visa Fraudster Applied ChatGPT to Trick Laiba Ejaz. Researchers Want Independent AI Safety Testing.

White Circle argues that AI firms should not be solely responsible for testing their own models.

For context, the organization is calling for mandatory third-party safety evaluations and notes independent researchers should be able to publish notable findings. AI-Generated Abuse Is Increasing.

For context, the findings come against the backdrop of growing concern concerning AI-generated non-consensual intimate imagery and child sexual abuse material.

Meanwhile, the Internet Watch Foundation remarked it identified 6, 310 AI-generated images gathering the legal definition of CSAM in the first half of 2026, around 40% more than the total discovered during all of 2025.

Researchers argue that stronger safeguards need to be built directly into AI models to prevent harmful content from being generated in the first place. Stay Connected with ProPakistani.

Obtain the latest tech news, telecom insights, and product launches wherever you prefer. Follow on Google Discover. Follow on Google News Join WhatsApp. See more ProPakistani stories in Google Search and Top Stories.

Technology and Automotive Specialist covering the latest cars, smartphones, AI breakthroughs, and.

Taken together, the developments around research Shows ChatGPT and Gemini Are Easily Tricked Into Making Nudes point to a situation that is still moving, and the coming days should bring more clarity.