How Dangerous Are AI-Built Viruses?

October 9, 2025 | 07:54 pm

TEMPO.CO, Jakarta - Scientists have begun using artificial intelligence (AI) to build new viruses that are different from existing strains or variants. This discovery opens up the possibility of creating AI-designed forms of life, while also raising concerns about the potential for misuse.

Researchers have developed bacteriophages, viruses that only attack bacteria. The research team has ensured that the AI model used cannot create viruses capable of infecting humans, animals, or plants.

Another study published in the journal Science on October 2, 2025, indicates that AI is also capable of bypassing security measures designed to prevent the development of biological weapons. In the study, researchers from Microsoft found that the AI system could circumvent safety mechanisms that would prevent bad actors from ordering hazardous chemicals.

Upon discovering this vulnerability, the research team promptly developed software patches to reduce the associated risks. However, the use of this system still requires high technical expertise with limited access.

A doctoral student at Stanford University, Sam King, has also conducted research on AI's ability to design new bacteriophages to combat bacteria causing infections. King's research, conducted with his supervisor Brian Hie, an assistant professor of chemical engineering, was published in the bioRxiv database in September 2025, although it has not yet undergone peer review.

"The state of this method right now is that it's quite challenging and requires a lot of expertise and time," said King, quoted from Live Science on Tuesday, October 7, 2025. "We feel that this doesn't currently lower the barrier to any more dangerous applications."

Security measures in the research include removing human-infecting virus data from AI training and conducting additional tests to ensure the model does not produce dangerous genetic sequences.

AI Virus Requires Supervision

However, some experts believe that security regulations in this field are not keeping pace with the rapid technological advancements. Tina Hernandez-Boussard, a professor of medicine at the Stanford University School of Medicine who consulted on the research security, warned that AI can breach safety boundaries if not tightly controlled.

"You have to remember that these models are built to have the highest performance, so once they're given training data, they can override safeguards," she said.

Several countries are taking anticipatory measures. The United States government, through executive orders in 2023, emphasized the importance of evaluating AI security and risk mitigation policies. In the UK, the AI Security Institute is also developing standards to prevent the misuse of this technology.

According to King, although AI theoretically has the capability to design new genomes to create life forms, there is currently no easy method to transform these designs into living organisms in a laboratory.

"There's a role for funders, for publishers, for industry, for academics," said Hernandez-Boussard. "This multidisciplinary community requires these safety evaluations."

Editor’s Choice: OpenAI Launches ChatGPT Pulse, a Personal Assistant for Summarization

Click here to get the latest news updates from Tempo on Google News

IISc Showcases India's Scientific Prowess and Expanding Global Collaborations

3 hari lalu

University of Indonesia to Open Artificial Intelligence Program Next Year

5 hari lalu

OpenAI to Block Users Generating Child Abuse, Exploitation Content

8 hari lalu

OpenAI Launches ChatGPT Pulse, a Personal Assistant for Summarization

10 hari lalu

China Proposes AI+ Global Cooperation Initiative at UN Meeting

13 hari lalu

Spotify to Introduce New Policy on AI-Generated Music

13 hari lalu

Vietnam's Hanoi Expands Public Services with Mobile Offices Across the City

13 hari lalu

Nvidia Develops New AI Chips Tailored for Chinese Market

14 hari lalu

Google Dismisses Claims of AI Overviews Harming Website Traffic

16 hari lalu

Danantara CEO Confirms Oracle's Investment Plans in Indonesia

17 hari lalu