Articles of Interest - Week 1/20 - 1/26
- Walker Robinson
- Jan 26
- 4 min read
Matt O’Brien, AP News. January 22nd, 2025.
Hours into his return to the White House, President Trump repealed the Biden administration's 2023 AI executive order that we covered in Week 10/21-10/27. While this action fulfills a campaign promise, many of Biden's AI initiatives have already been completed through agency reports and recommendations. Furthermore, some key provisions remain in place, including the recent order on AI data centers. The tech industry's response has been measured for the most part. Even Trump supporters like Scale CEO Alexandr Wang declined to criticize specific aspects of Biden's order while expressing optimism about future collaboration with the new administration.
Despite the symbolic repeal, there's significant continuity in U.S. AI policy. Much of Biden's order actually built upon initiatives that Trump started during his first term. The most concrete change appears to be the removal of requirements for companies to share safety test results of powerful AI systems with the government. I worry this shift away from federal oversight may lead to far more significant risks in AI advancement in the future. That said, Trump's positive response to Biden's recent order on AI data centers and his support for the $500 billion Stargate project suggests some pragmatic considerations might outweigh political messaging regarding building AI infrastructure and continuing to lead the world in AI technology. It is difficult to predict where the Trump Administration will head with AI. Still, one thing is for sure: AI advancement will not stop, so the U.S. government cannot afford to disregard it or fail to pass the necessary policy needed to ensure we benefit from this technology and are not threatened by it.
John P. Mello Jr, TechNewsWorld. January 21st, 2025.
Cisco announced a new security "solution" for protecting companies' AI systems from emerging threats in an era of rapid technological development. Cisco's "AI Defense" solution is intended to help companies defend against new and evolving threats to their AI systems, especially ones that more traditional cybersecurity may struggle to deal with. The AI Defense platform offers features including detection of shadow AI applications, automated testing for safety and security issues, and continuous validation against threats like prompt injection and data leakage. Industry experts note that while Cisco's extensive networking infrastructure puts it in a unique position to provide such protection, the solution might have some limitation issues. The more complex AI security challenges (especially at the training stage) could prove too difficult for companies like Cisco to protect against.
There is a growing disconnect between current cybersecurity approaches and the unique challenges posed by AI systems. The emphasis on protecting against "shadow AI" leads me to think that organizations are already losing control over their AI deployments. If that's true, it would be highly concerning, to say the least. The timing of this release, as major labs push toward AGI, adds urgency to these concerns. If the U.S. and China (or companies within) are going to pursue a race to AGI, the security and safety aspects must be sufficiently addressed. Cisco's solution is certainly a step in the right direction. Still, there will likely be some limitations to the areas of the AI development process that the solution is able to sufficiently defend. Securing the AI development lifecycle and training pipeline will likely be challenging. In my opinion, we will likely need a more comprehensive, national-level approach to AI security that goes beyond what any single company can provide.
Makoto Tsujiguchi, Moyuru Tanaka, CSIS. January 21st, 2025.
A new CSIS analysis highlights emerging opportunities for defense technology collaboration between U.S. and Japanese start-ups. While the U.S. accounts for 42% of global arms exports, Japan has historically had minimal defense exports due to longstanding restrictions following World War II. However, recent changes are creating new opportunities for bilateral cooperation. These changes include Japan's ¥1 trillion investment in start-up ecosystems and a new security clearance system. The article discussed the Pentagon's plans to source "dual-use" technology from allies worldwide. Japan is an ideal partner for the U.S. to source emerging technology from and to pursue R&D cooperation with. Japan's advanced technology ecosystem is substantial and offers numerous opportunities for the DoD to pursue technological acquisitions. Furthermore, the U.S. and Japan have a preexisting and well-established trade relationship, making new endeavors easier. The analysis ends by recommending a few paths forward for the two countries. First, the U.S. and Japan should involve start-ups in the dialogue concerning emerging technologies and military use. Second, Japan must improve its information security to protect sensitive information related to emerging technology from adversaries like China. Third, the authors state that Japan's start-ups must move away from an academic mindset that has led to their hesitancy to work with the military. Lastly, Japan should better enable its small and growing start-ups to pursue funding and connect with U.S. investments.
This transformation of the U.S.-Japan defense relationship reflects growing recognition that commercial technology from start-ups will likely be crucial to the future of military technology and warfighting. The depth of proposed cooperation is thorough, but also precisely the type of cooperation that the U.S. and its allies should be engaging in to prepare for a future that will contain extremely high levels of technology, especially in military conflicts. The potential results could significantly enhance alliance capabilities and create an established norm for emerging technology information sharing and joint-development. If successful, this initiative could serve as a model for similar agreements and partnerships with other U.S. allies. Creating a network of like-minded, allied, and technologically advanced nations that bolster each other's technological innovation and national security would enable greater capacity to compete with adversarial nations like China and lay the groundwork for tackling the challenges and threats accompanying emerging technologies. An already established communication and information-sharing network would go a long way when dealing with crisis situations and damage mitigation should something terrible happen with an emerging technology.
