Anduril Founder Luck on AI Cybersecurity: US AI Offense-Defense Imbalance, and in the Long Run the Side That Uses AI Well for Defense Will Hold the Advantage
IT之家 November 19 news, Anduril co-founder 帕尔默 · 拉奇 (Palmer Luckey) today at The New York Times DealBook Summit warned that the US is facing a severe imbalance in its AI cyber offense and defense systems.
拉奇 said AI can already help attackers develop various exploit programs to launch attacks against US enterprises and infrastructure. "Attackers rapidly adopt new technologies, while legacy systems cannot complete upgrades as quickly as technology develops—there is a mismatch between the two."
He believes the US needs to accelerate the renewal of digital infrastructure and the building of security protections. "I think everyone must pick up the pace; it is time to start accelerating."

The remarks came as several US institutions have recently experienced successive technical failures. Luck linked these incidents to the enhanced cyberattack capabilities driven by AI technology (however, there is currently no evidence that these failures were caused by cyberattacks). 拉奇 believes the related attacks may be for strategic purposes or for economic gain.
拉奇 believes AI will not only strengthen attack capabilities but may also give defenders a greater advantage. He said that in the long run, the side that can most effectively use AI for defense will hold the advantage, but currently attackers are ahead of many traditional defense systems in the speed of technology adoption.
Beyond cybersecurity, Luck also discussed the development of US defense technology. Anduril, which he founded, mainly develops AI drones, autonomous weapon systems, and military software. US Secretary of Defense 皮特 · 赫格塞斯 previously invited Luck to participate in a 120-day strategic study called Project Meridian to study the direction of future warfare. Luck defended this arrangement, saying there is a long tradition of government and industry figures discussing the direction of military development together.
Luck also disclosed a maritime construction plan he is advancing, SLAB. According to information he posted on X, the project hopes to build large offshore platforms in US coastal waters for industries such as metal processing, oil refining, chemical engineering, medical research, and seawater desalination, lowering land and operating costs while reducing state and local regulatory restrictions.

拉奇 said the platform's design life target is 1000 years, the project is jointly invested by him and several friends, and as early as 2017 it received funding support at a valuation of 1 billion US dollars (IT之家 note: at current exchange rates about 6.681 billion yuan). He said the project still has a long way to go and was not proposed recently due to AI data center demand.
Also worth mentioning is that 拉奇's own entrepreneurial experience also became one of the topics discussed at the summit. In 2014, the virtual reality company he co-founded, Oculus VR, was acquired by Facebook for about 2 billion US dollars (at current exchange rates about 13.362 billion yuan). He then joined Facebook, but later had disagreements with the team over the direction of the VR business.
Before the 2016 US presidential election, 拉奇 donated 9,000 US dollars to an organization supporting 特朗普. The organization had published attack ads targeting Democratic presidential candidate 希拉里 Clinton. This incident caused dissatisfaction among Facebook employees; 拉奇 said colleagues demanded the company fire him, and he eventually left in 2017.
Speaking about this experience, 拉奇 said he is still bitter about it to this day. He said: "I'm still very upset about it. I have never let it go." 拉奇 also described founding Anduril as an important entrepreneurial project after leaving Facebook. Today, the company has become an important player in the US defense technology field, with business covering autonomous unmanned systems, artificial intelligence, and military equipment manufacturing.
Related Articles
- Marsh: There are signs hackers are using AI tools and agents to find security vulnerabilities
- Anduril and Qualcomm team up to launch the "Made in America" laptop BlueChip, promising that when it goes on sale next year the share of domestic components will reach 65%
- Google's strongest AI model: Gemini 4 Carbon tops the global intelligence index, "dedicated" to cybersecurity
- Qualcomm and Anduril launch the BlueChip laptop: about 90% of components come from the US and its allies, targeting military and government customers
- OpenAI, Microsoft, Google and 116 other companies issue a joint letter calling for high attention to cybersecurity in the AI era
- With frequent cyberattacks, French government departments will use AI to detect cybersecurity vulnerabilities