4 Indian researchers develop unique technology in the United States to prevent ID leakage during AI photo editing – everything is related to “PrivateEdit”

The researchers disclosed the project in a conversation with HindustanTimes.com. Talking about the inspiration behind the project, Dipesh explained that it “came from a specific moment in early 2025, when AI “Ghibli-esque” filters Go viral. “

viral trend Ghibli style portraits It took social media by storm last week as netizens asked artificial intelligence to alter their photos into a unique anime style. Although many people like New features in OpenAIsome slammed the trend as “disrespectful” to the Studio Ghibli co-founder Hayao MiyazakiHe has previously criticized AI-generated animations as “an affront to life itself” and said he “doesn’t want to incorporate this technology into my work at all.”

As new technology emerges, people may wonder what never-before-seen capabilities it can offer. Dipesh has an answer for this.

Learn more | OpenAI, Anthropic, and Google join forces to crack down on China’s artificial intelligence model copying

“Most privacy tools today are ‘reactive’ – they try to fix problems after your data has been sent. PrivateEdit is ‘privacy by design’. We introduce a way to ‘decouple’ your identity from the rest of the image. The real novelty is that our technology works with the large AI models you’re already using, like Midjourney or ChatGPT) without requiring them to make any changes. We’ve also introduced a “Trust Slider” to put the power back into the user’s hands; you can decide exactly how much information to hide based on how much you trust a particular platform, a level of personalization that didn’t exist until now.”

“We developed a pipeline that acts as a ‘safety filter’ between you and the AI. Instead of sending your entire photo to the cloud, our system first runs locally on your device,” he told us hindustan times.com. “It uses advanced segmentation to find the “identity-sensitive” parts of your face (that make you your unique mark) and puts a digital mask over them. We then send only the “background” and masked versions to the AI. The AI ​​performs the edits you requested and then sends the photo back to your device, where your real facial details are securely reinserted. The AI ​​does its job, but it never actually “sees” the real you.”

One might wonder if this technology is only for tech experts, or if it can also be used by regular smartphone users. Dipesh said they were determined not to make this just a “lab experiment” and stressed that it was user-friendly.

“The goal is to make it feel like a standard photo editing app. You don’t need to know how AI works or what ‘split’ is; you just use a simple slider to select your privacy level, and the app handles the complex ‘masking’ and ‘rebuilding’ in the background. Privacy shouldn’t be a chore; it should be as easy as applying a filter,” he said.

“Many users believe that their photos will be deleted once a ‘filter’ is applied, but this data often becomes part of a permanent digital footprint, used for surveillance, analysis or training future models without explicit consent. In the current scenario, your biometric identity is being acquired as a commodity. Moving to a ‘design-to-privacy’ framework like the one we have developed is critical to ensuring that the AI ​​revolution does not come at the expense of basic human autonomy,” Vaneet explains. Vinit worked.

Learn more | Economist PhD says his work now involves removing the ‘dash’ from AI writing: ‘The invisible labor of the AI ​​economy’

Technical collaborator Ataf explains what risks the Purdue researchers’ new technology could reduce.

“When you upload raw photos, they can be stored indefinitely, leaked in server vulnerabilities, and even used to train ‘deepfakes’ without your permission. By using our shielding system, sensitive data is never even transmitted to the cloud. This also helps companies; they can now offer AI photo tools to customers without the huge legal and ethical responsibility of storing thousands of people’s private facial data,” he said.

Aharv also claimed that the technology could be used by big companies such as Adobe, Apple or Google, calling it the “ideal future of this technology.”

“Because our pipeline doesn’t require companies to change their existing AI models, it can be integrated into current applications as a ‘privacy layer.’ This will allow these big tech companies to offer amazing generation capabilities while proudly telling their users: ‘We’ve never even seen your original photo.’ “It’s a win-win for the company’s reputation and the safety of its users,” he explains.

“Most laws focus on what companies do after they get the data. Our work provides a technical path to ‘data minimization’, a key principle of privacy laws such as GDPR. By demonstrating that we can achieve high-quality results without collecting sensitive data, we are providing a blueprint for how future AI regulations will be developed,” he explained.

Vineet reveals that the biggest challenge is that covering the face can make the final AI-edited photo look fake or affect its quality.

“If you cover it too much, the AI ​​will lose context and the photo will look weird. If you cover it too little, privacy will be compromised. We developed a ‘smart blending’ technology that gives the AI ​​enough information to understand the lights and shadows of a scene without seeing your actual biometric features. The result is a high-quality, professional image where the ‘seam’ between your real face and the AI ​​edit is completely invisible,” he said.

At the same time, Dipesh said one thing people must remember when using AI tools in the future is that “innovation doesn’t have to come at the expense of your identity.”

“For too long we have been told that to get the best technology we have to give up our data. Our research proves this is not true. You can have the most powerful AI in the world and your privacy. You will never have to choose between creativity and security,” he concluded, adding that the next important step is verifiable data sovereignty.

“It’s not enough for companies to promise not to use your data; we need technological systems where users can mathematically verify that their data is only used for the tasks they requested and then deleted immediately. Combining this with on-device processing will be key to an AI world where innovation and personal security are not in conflict,” Dipesh said.