Prevent Data Leaks, Telkom Develops Internal Chatbot for Employees

JAKARTA - PT Telkom Indonesia is building an internal chatbot platform based on artificial intelligence (AI) as a proactive step against the widespread use of public AI such as OpenAI through its product ChatGPT as well as Gemini and similar applications.

Telkom Indonesia's Executive General Manager of Digital Products, Komang Budi Aryasa, said the initiative was born from the phenomenon of public use of AI by employees who were considered practical but had the potential to pose data security risks.

"It was the first time when the company didn't have a platform like ChatGPT internally, employees tended to use ChatGPT because it was easy," Komang said at the Telkom Ramadan gathering, in Jakarta, written on Tuesday, March 3.

According to Komang, the ease of using public chatbots often makes employees miss out on potential risks, especially when they unknowingly upload company internal documents to external platforms.

"When we upload the document, the document is stored on their server. It could be our competitors, it could be irresponsible people, then the data leak occurs," he said.

To anticipate this risk, Komang emphasized that companies need to take firm steps from top to bottom.

Management, he said, must issue official rules prohibiting the use of public AI for employment purposes.

"Companies have to issue enforcement, so rules. If necessary, there will be a punishment if they are found to have submitted company documents to this open application," he said.

Komang said that a number of companies had even blocked access to public AI applications through internal intranet networks.

If employees need AI assistance, they are directed to use the company's built-in GPT.

"If you want help, just use the internal GPT that we have built, it's the same result," said Komang.

Komang admitted that the level of maturity of the use of AI in Indonesia is still at an early stage and is developing gradually.

Therefore, continued Komang, Telkom considers it important to build awareness (awareness) regarding the use of AI in increasing productivity and competitiveness of companies.

"Well, the adoption is actually when we implement AI in some companies it is quite decent," he said.

Komang said, specifically for chat or chatbot solutions, the main challenge lies in the need for training and periodic knowledge updates.

Chatbots, continued Komang, must continue to be given new data and knowledge so that they are more accurate and less error-prone.

"What we ask sometimes the answer is different, it's a hallucination so it needs additional knowledge. So when we deliver lines and chats, new knowledge is still needed so that they get smarter and smarter," said Komang.