A former U.S. national security official has argued that, to prepare for the possibility that China develops artificial general intelligence (AGI) before the United States, Washington should be ready not only for cyberattacks but also for physical strikes on data centres.

On Sept. 11 local time, IT outlet TechRadar reported that Jacob Stokes, a former national security official from the Barack Obama administration, made the argument in a document titled “Superpowers and AGI,” saying the United States should review in advance the tools it could use to stop China if it secures AGI first.

The options Stokes presented included cyberattacks and sabotage, as well as physical attacks to destroy Chinese AI data centres with missiles or bombs. Viewing China’s AGI development as a direct threat to U.S. security, he stressed that a national-level response strategy is needed beyond a technology race.

AGI refers to AI with general reasoning and problem-solving ability at a level similar to, or beyond, humans. Stokes voiced concern that AGI could show self-preservation tendencies or be used for large-scale cyberattacks and development of new weapons. He in particular presented a scenario in which China could secure AGI first through an unexpected technological breakthrough.

He said response tools should be reviewed in stages. At a relatively low level, it could use cyberattacks targeting AGI systems, and if attacks fail or the threat level rises it could move to stronger measures. He assessed physical attacks as the most dangerous option with the highest risk of escalation.

The problem is that such a response could lead to actual military conflict. Direct attacks on China’s data centres could trigger Chinese retaliation, and uncertainty is high, from judging whether China has actually secured AGI to judging which country could operate the technology more responsibly.

Ultimately, the argument shows that the AGI race may not end as a technology contest to build better AI first. If the United States and China regard AGI as a national security asset, AI data centres and computing infrastructure are also more likely to be treated as strategic facilities.

Still, with neither the timing of AGI’s arrival nor its actual capabilities confirmed, response strategies that include pre-emptive physical attacks are expected to generate significant controversy. In the future AGI race, key tasks are likely to include not only who secures the technology first but also how to verify its development and how to set up deterrence and verification systems to prevent miscalculation and escalation.