DICT, DOST warn public vs AI chatbots

MANILA, Philippines — The Department of Information and Communications Technology (DICT) and Department of Science and Technology (DOST) are urging Filipino cyber geeks and netizens to be wary of engaging with artificial intelligence chatbots, following worldwide disruption caused by a Chinese AI startup.

DICT Secretary Ivan John Uy said that people should be “more discerning” in engaging with AI chatbots regardless of who developed the bot, especially with the details these machines disclose.

“As you make your queries, as you engage with AI, you will be sharing information, you will be sharing data. So you have to be careful as to what extent (of your personal data) you want to share,” Uy said at the Kapihan sa Manila Bay breakfast forum in Malate, Manila last week.

According to Uy, the DICT was closely monitoring the disruption in the global ICT and financial industries last week, caused by Chinese AI startup DeepSeek’s launch of their DeepSeek-R1 large language model. According to analysts, DeepSeek -R1 holds its own against OpenAI’s ChatGPT, if not better.

The arrival of DeepSeek-R1 caused a one-day crash in the US tech market, with graphics processing unit (GPU) and microchip giant Nvidia the most hit. Meanwhile, other companies, such as Broadcom, Taiwan Semiconductor Manufacturing Co., Microsoft, Meta and even Amazon were also affected.

“This DeepSeek is disruptive and game changing because it shattered the preconceived notion that one needs huge amounts of GPUs and hundreds of millions of dollars to develop AI (large language models),” Uy said.

However, the ICT secretary stressed that any generative AI bot or platform needs data to attain high intelligence level.

“Data is the fuel of AI. AI is the engine, but AI won’t run without fuel, and data is the fuel of AI. The more data you feed the AI, the more intelligent the AI will be. The less data you provide, the more limited the AI’s capability will be,” Uy said.

“So fine, you can share data, but be more discerning on what kind of data you share, because it will be harvested by the AI, and they will add it to their collection of data and you cannot get it back,” Uy said.

Meanwhile, a well-informed source at the DOST said that initial evaluation of DeepSeek R1 has shown an apparent geopolitical bias toward China.

The source said that they have already detected limitations on the DeepSeek AI that apparently prevent it from answering certain queries about Chinese leader Xi Jinping and the West Philippine Sea.

These early findings, the source said, raise the need for further evaluation of DeepSeek not just in terms of its capabilities but also on security.