‘The control is totally with users’: Samsung SRI-Noida MD on privacy in the age of Agentic AI

As artificial intelligence (AI) shifts from reactive tools to proactive “agents,” Samsung is drawing a firm line when it comes to user privacy. According to Kyungyun Roo, managing director, Samsung R&D Institute of Noida (SRI-Noida), the tech giant is prioritising a “user-first” philosophy, ensuring that AI becomes more capable but never intrusive.

The South Korean company launched the Galaxy S26 series as an ‘Agentic AI’ phone where the underlying technology becomes more capable and works in the background, enabling users to provide relevant responses when they are communicating on their phones. One of the Agentic AI features on the latest flagship smartphone series is Now Nudge that recognises what is shown on the screen and suggests relevant actions, such as adding events to your calendar or sharing photos.

The core mission of Agentic AI on Samsung phones, according to Roo, is to collapse complex digital chores into single-step actions. Speaking on the development of the Now Nudge feature, Roo explained that the goal is pure efficiency. “Now Nudge generally solves the productivity in terms of efficiency, creating the multiple steps into one step,” Roo told The Times of India tech. He illustrated this with a common social scenario:“For example, we have been to a party, and you chat with me asking for photos of it.

So, Now Nudge will be prompting the suggestion that these are the photos respective to you and me. So, the user can save the multiple clips and share afterwards” (instantly) – without having to hunt for those photos through their entire library.Roo said that SRI-Noida’s framework team worked on this feature. The centre was established in 2007, and the major products that the centre works on are Galaxy devices such as smartphones, tablets and wearables (Galaxy Watch and Buds). “We started from feature phone development, moved to smartphones, and now we’re working on AI phone development. Our R&D center is young and energetic and we work on three key pillars – innovation, expertise build-up and strategic partnership,” Roo said.

Samsung on how it decides on the ‘nudge’ threshold

One of the biggest challenges for Samsung’s R&D centre is calibrating the ‘threshold’ – the invisible boundary where a helpful suggestion turns into a digital nuisance.

Roo explained that in order to prevent the system from becoming overbearing, the framework team focuses on context. This way, Samsung explained, the AI only offers suggestions when they are directly relevant to the task at hand, acting as a supportive assistant rather than an uninvited guest.However, for many users, the idea of an AI “scanning” a photo gallery raises immediate red flags. Samsung addressed these concerns, clarifying that the AI operates under strict user-defined boundaries.

When asked how Samsung sets the limit on the level access of the feature to, for example, users don’t want certain photos to be accessed by AI, Roo noted that the AI cannot “see” anything until a user explicitly grants permission for it to access the Gallery.“The control is totally with users. First, they have to provide the permission for Gallery, and then only AI will access photos. Based on the conversation content, pictures as per user permission only will be suggested,” Roo explained.

He also highlighted that for photos that the user does not want to be accessed, Samsung offers other methods. For example, moving pictures to the private album, which is a “secret” album.While talking about AI’s limitations in following commands such removing the clothes or adding nudity, Roo said whenever Samsung launches a feature, it is perfectly safe. “There are lots and lots of additional algorithm that we put just to ensure that inappropriate contents are not shown. There are many features even in the Gallery, where we do not show certain type of generative contents or answer inappropriate questions,” Roo said.