Gladstone.AI Announces the First-Ever AI Action Plan for United States National Security Commissioned by the U.S. State Department

News provided by

Mar 11, 2024, 12:00 ET

The report, "Defense in Depth: An Action Plan to Increase the Safety and Security of Advanced AI" outlines a government-wide framework to address risks associated with the rapid development of advanced AI on the path to AGI.

This first-of-its-kind government-commissioned report consolidates input from over 200 stakeholders across the U.S. national security community and industry, and includes an analysis of the competitive dynamics among AI labs, including a novel and direct perspective from frontier lab AI researchers who have privately voiced AI safety and security concerns.
The report includes a comprehensive Action Plan to address catastrophic national security risks from advanced AI on the path to AGI. The Action Plan recommends interim measures, early warnings and contingency planning, investments in AI safety and security research, legal safeguards, and international capacity-building — all to support the responsible development and adoption of advanced AI in the face of potential catastrophic threats to national and global security.

WASHINGTON, March 11, 2024 /PRNewswire/ -- Today, Gladstone.AI, the leader in safety-forward AI development and national security policy, unveiled a first-of-its-kind report, "Defense in Depth: An Action Plan to Increase the Safety and Security of Advanced AI", which was delivered as part of an assessment commissioned by the United States Department of State.

The release of "Defense in Depth" comes at a pivotal moment, in which the United States is coming to terms with the double-edged sword of advanced AI development: the promise of transformative opportunities alongside the peril of new types of catastrophic risks. The Action Plan's recommendations carve out an urgently needed path to stabilize competitive race dynamics in AI, strengthen our national capability and capacity, and scale safe and secure AI development. It offers a safety-forward approach to rapid development alongside safeguards that protect our national security interests.

"The Action Plan bridges the gap between Silicon Valley and Washington in a way no previous initiative has ever attempted," said Jeremie Harris, CEO and Co-Founder of Gladstone AI. "The entire assessment was built to meet policymakers where they are with a practical, technically informed, data grounded roadmap to stabilize, strengthen, and scale AI development safely."

"'Move fast and break things' can be the right approach when you're building low-stakes systems and your downside is capped; but when your downside is significant, the responsible thing to do is adopt a safety-forward development approach. This is how the Department of Defense develops their own high-stakes systems, and it's a philosophy we'll need to start applying to advanced AI as systems become increasingly capable," said Edouard Harris, Gladstone AI's CTO and Co-Founder. "We're proud to advance a new kind of safety-forward AI policy framework."

Action Plan Recommendations // The Action Plan advocates for a "Defense in Depth" strategy, in which multiple mutually supporting controls guard against any single point of failure. It outlines a proactive and technically-grounded approach to mitigate potential catastrophic risks at this pivotal moment in AI development. The Action Plan lays out five Lines of Effort, which together put us on a path to stabilize, strengthen, and scale advanced AI development safely and securely:

Stabilize the current race to AGI: address near-term risks by establishing baseline safeguards, including export controls and monitoring, just like we would for other societal harms like toxic chemicals or bioweapon hazards.
Strengthen our capabilities: expand our preparedness and response capabilities through early warning systems, contingency planning, and investment in technical AI safeguards research.
Scale safely: formalize long-term AI regulatory and legal safeguards, and lead internationally, economically, and strategically to protect and further our interests globally.

About the project // The "Defense in Depth" Action Plan is the product of discussions over 13 months with more than 200 stakeholders, from CEOs and other executives at leading AI labs, to researchers, U.S. government WMD experts, and some of the most senior national security officials in the United States and around the world, culminating in a first of its kind AI Action Plan for U.S. national security. It is a landmark effort to assess and address catastrophic risks from the rapid development of advanced AI on the path to AGI. Through a combination of historical analysis, technical research, stakeholder engagement, and expert policy recommendations, the plan charts a course for mitigating the risks of weaponization and loss of control that AI may soon introduce. The research behind the action plan underscores the importance of a proactive, informed approach to AI safeguards, drawing on lessons from the history of nonproliferation to navigate future challenges. To request a copy of the full Action Plan, and for summaries of its recommendations, please go to https://www.gladstone.ai/action-plan.

About Gladstone AI // Gladstone AI was founded by national security experts and experienced Silicon Valley AI executives. The company is dedicated to helping the U.S. national security and defense communities address the opportunities and risks that come with unprecedented and rapidly accelerating AI capabilities. Gladstone is an inaugural member of the Department of Commerce's first AI Safety Consortium (AISIC), and offers a combination of AI training programs, AI-powered products, and consulting services for national security agencies aiming to prepare themselves to address the challenges of the new era of AI.