Georgia Tech researcher examines AI safety in IBM internship

Georgia Tech Ph.D. student Mansi Phute spent a 2026 internship at IBM examining the security of communication between multimodal artificial-intelligence agents, according to reporting by Newswise. The placement gave her an opportunity to continue working in AI safety while seeing how research priorities are shaped inside a technology company.

Phute is a researcher in machine-learning robustness and AI safety at Georgia Tech, where she is advised by Polo Chau, a professor and associate chair in the School of Computer Science. She completed the internship with IBM’s AI Foundations team in San Jose, California, before returning to the university.

The experience offered a direct comparison between two environments that often approach scientific work differently. Universities commonly emphasize fundamental knowledge and long-term inquiry, while companies must also consider how research will be implemented and connected to wider organizational goals. For Phute, the internship provided an opportunity to understand both perspectives without moving away from her thesis interests.

A focus on multimodal agents

Phute chose IBM because the company has a strong presence in AI safety and supports exploratory research, Newswise reported. She expected the internship to expose her to advanced industrial approaches while allowing her to continue developing as an academic researcher.

Her work centered on the safety of communication among multimodal agents. These systems can work with more than one type of information, and Phute described the issue as increasingly important as agents are used in a growing range of applications. The subject also closely matched her research interests in making machine-learning systems more robust and artificial intelligence safer.

The internship built on preparation from Georgia Tech. Phute said her coursework and other academic experiences transferred effectively to the IBM setting. She also emphasized the value of working with people from different research backgrounds during her studies. Familiarity with multiple areas of research helped her communicate with colleagues and adjust to a new professional environment.

That ability to move between fields became especially useful in an industry setting, where projects can involve technical questions alongside safety, policy and alignment concerns. Conversations with colleagues working in those areas helped Phute compare the way industry frames AI safety with the more academically oriented perspective she had previously encountered.

Learning to communicate across audiences

One of Phute’s most significant moments came when she presented her work to an audience that included both technical and non-technical staff from several teams. After the presentation, her manager praised her ability to explain research to people with different levels of familiarity with the subject.

The response was meaningful to Phute because she had intentionally worked to improve her communication. Presenting complex research clearly to varied audiences was not separate from the technical work; it became part of what she learned about conducting research in a large organization.

The internship also showed her how AI safety projects relate to a company’s broader business priorities. In an academic setting, it can be difficult to see how a research question connects to product development, deployment or organizational decision-making. At IBM, Phute was able to observe those connections directly while working with experienced researchers in her field.

Why industry matters to AI research

Phute returned to Georgia Tech with a stronger view of the role private companies can play in scientific progress, particularly in artificial intelligence. She pointed to industry access to computing resources, which can be less restrictive than the resources available in academic environments. That access can support research that may otherwise be difficult to pursue at scale.

She also sees value in understanding how companies implement and deploy research. Awareness of industry priorities can help researchers shape questions and methods that remain scientifically meaningful while also contributing to practical applications. In AI safety and reliability, she believes this connection can increase the effect research has beyond the academic community.

The internship did not replace Phute’s academic focus; it broadened it. By working with IBM’s AI Foundations team and discussing safety, policy and alignment with specialists, she gained a clearer picture of how private industry approaches problems that are also central to her university research.

Newswise’s account presents the placement as an example of the growing interaction between academic investigation and industrial development in AI. For Phute, that interaction supplied both technical experience and a better understanding of how research can move from foundational ideas toward systems used in the real world.