AI Safety’s Power Play: CAIS Taps xAI Talent, Targets National Security

SAN FRANCISCO, CA – June 02, 2026 – The Center for AI Safety (CAIS), a nonprofit that has quietly been building a reputation for rigorous technical research, has just made its most assertive move yet. In a dual announcement that reverberated from Silicon Valley to the Capital Beltway, the organization revealed it has appointed former xAI leader Devin Kim as its first President and is simultaneously launching the Frontier Security Institute (FSI), a new Washington, D.C.-based entity aimed squarely at the national security apparatus.

This strategic expansion is more than a simple leadership shuffle; it represents a significant pivot for CAIS and a telling indicator of the maturation of the AI safety field. The organization is signaling a deliberate shift from a primarily research-focused body to an operational powerhouse determined to embed itself at the critical intersection of frontier AI development and U.S. national security policy. By bringing in top-tier talent from the heart of AI development and establishing a permanent foothold in D.C., CAIS is positioning itself as an indispensable arbiter in an era where algorithms are becoming instruments of state power.

"Today marks a significant moment for CAIS and for the field," said Dan Hendrycks, Executive Director of CAIS, in a statement. "Frontier AI is now a national security technology, and the National Security Enterprise needs partners fluent in both worlds. We are building the organization and team that can do that work."

A New Guard for AI Safety

The appointment of Devin Kim as President is a strategic coup. Kim is not a policy theorist or an academic; he is a builder who has operated at the bleeding edge of AI development. As an early employee at xAI, he led post-training and research infrastructure for the Grok models, Elon Musk's answer to competitors like OpenAI. Before that, at Scale AI, he focused on trust and safety systems, working directly with teams at OpenAI and Meta to fine-tune model performance and safety.

His transition from a top-tier AI lab to a safety-focused nonprofit exemplifies a growing trend: the migration of elite technical talent towards governance and risk mitigation. As AI models become more powerful and unpredictable, the engineers who know their inner workings best are increasingly recognizing the imperative to build guardrails. Kim’s move lends significant credibility to CAIS, suggesting that the most critical work is no longer just about advancing capabilities, but about controlling them.

"I have spent my career building the most powerful AI systems in the world and am a firm believer that AI has the potential to profoundly benefit society," Kim stated. "Realizing that potential depends on getting safety right, and CAIS is where that work happens."

In his new role, Kim will oversee CAIS’s research mission and organizational strategy, but his most crucial function will be translation. As Hendrycks noted, "Devin has built systems at the frontier of AI development, understands the research infrastructure that makes safety work possible, and can translate all of it for audiences outside the technical community." This ability to speak both languages—that of the San Francisco-based engineer and the D.C.-based policymaker—is the core asset CAIS is betting on.

Forging the National Security Bridge

If Kim’s appointment bolsters CAIS’s technical credibility, the launch of the Frontier Security Institute (FSI) solidifies its strategic ambition. Based in Washington, D.C., FSI is explicitly designed to act as the “translation layer between frontier AI developers and the national security operators who must acquire, govern, and utilize these systems.” Its mission is to close the widening capability gap between the private labs racing to build ever-more-powerful models and the government agencies tasked with protecting the nation.

The institute’s leadership is a clear signal of its intent. At the helm as Executive Director is Isaac "Ike" Harris, a 23-year U.S. Navy veteran who commanded a destroyer and later served as a policy advisor to the Secretary of Defense on China and technology security, as well as a senior staffer on the House Select Committee on the Chinese Communist Party. His deep experience within the Pentagon and on Capitol Hill provides FSI with immediate access and institutional fluency.

"As the labs race to expand the frontiers of AI capability, the National Security Enterprise risks falling behind," Harris warned. "Bridging the two has never been more critical to American national security or to the future of American innovation."

Joining Harris is Chief Operating Officer Jeremy Pelter, whose nearly two decades in federal service include stints as Acting United States Secretary of Commerce and Acting Under Secretary for Industry and Security. This pairing of a seasoned military strategist with a high-level administrative expert gives FSI a formidable operational foundation. The institute's initial focus will be on issues unique to military and intelligence use cases: securing advanced models from theft or manipulation, establishing robust testing and evaluation protocols for defense applications, and analyzing how AI is reshaping geopolitical stability.

Navigating a Crowded and Urgent Landscape

CAIS is not entering an empty field. The national security implications of AI have become a top priority for the U.S. government and its allies. The FSI will have to navigate a complex ecosystem that includes the government's own U.S. AI Safety Institute (housed at NIST), which has already inked agreements with major labs for pre-deployment evaluations, and the National Security Agency’s Artificial Intelligence Security Center (AISC), tasked with defending the nation's AI capabilities.

Furthermore, established think tanks like Georgetown's Center for Security and Emerging Technology (CSET) and the RAND Corporation have been producing influential research on these topics for years. However, CAIS and FSI appear to be carving out a unique niche. While government bodies are focused on official oversight and think tanks on policy analysis, FSI aims to be a more agile, hands-on intermediary—a conduit for practical knowledge and rapid dialogue between the fast-moving tech sector and the more deliberative national security community.

The timing is critical. Recent executive orders have created pathways for federal vetting of AI models, and congressional committees are holding increasingly frequent hearings on the security risks of frontier AI. FSI's launch positions it to be a key resource and influential voice in these exact conversations, offering a blend of technical depth and security clearance that few other organizations can match.

This expansion transforms the Center for AI Safety from a respected research institution into a strategic player with national consequence. By recruiting a top mind from the AI frontier and building a dedicated bridge to the national security world, CAIS is making a bold claim: that in the 21st century, ensuring the resilience and permanence of our institutions requires mastering the very technology that threatens to disrupt them. The success of this venture will depend on its ability to maintain the trust of both the secretive AI labs and the security-conscious institutions it seeks to serve, a balancing act that will be essential for navigating the unpredictable landscape ahead.