Anthropic launches research institute to study AI risks while battling Pentagon blacklist

Anthropic Creates New Research Body Amid Government Tensions

Anthropichas launched the Anthropic Institute, a new internal think tank designed to examine the risks associated with artificial intelligence as the company simultaneously battles a Pentagon blacklist and expands its Washington presence 1

1

. The institute will focus on AI's large-scale implications, including what happens to jobs and economies, whether AI systems make society safer or introduce new dangers, and whether humans can retain maintaining control over AI systems 1

1

.

The timing is particularly notable given Anthropic's ongoing legal fight with the Defense Department over a supply chain risk designation that resulted in President Trump ordering federal agencies to stop using the company's technology 2

2

. Despite these challenges, the company is making aggressive moves to influence future AI policies through expanded operations in the nation's capital.

Leadership Reshuffling and Strategic Expansion

Co-founder Jack Clark has transitioned from his role as Head of Policy to become Head of Public Benefit, where he will lead the Anthropic Institute 2

2

. Sarah Heck, a former Stripe executive who recently joined as Head of External Affairs, has taken over Clark's previous position as Head of Public Policy 4

4

.

Anthropicis tripling its Public Policy office team and opening a permanent office in Washington D.C. this spring to engage policymakers and think tanks long term 3

3

. "AI is advancing faster than any technology in history, and the window to get policy right is closing," Heck stated, emphasizing that the company is building a bipartisan team to ensure smart policy can accelerate American innovation rather than slow it down 3

3

.

Three Teams Unite Under One Research Umbrella

The Anthropic Institute consolidates three existing research teams to study the societal impacts of AI comprehensively. The Frontier Red Team, responsible for stress-testing AI systems to understand their limits and capabilities, recently used Claude to scan Firefox's code base for vulnerabilities and tested whether AI can autonomously develop exploitation methods 4

4

3

3

.

The Societal Impacts team collects data on how users interact with Claude in real-world scenarios. Last month, it published research evaluating why and when workers allow AI agents to perform tasks autonomously 4

4

. The Economic Research team tracks the economic effects of AI on jobs and broader markets, publishing Anthropic's Economic Index report that details what business activities customers are automating with Claude 4

4

.

High-Profile Hires Signal Serious Ambitions

Anthropichas recruited significant talent to strengthen the institute's capabilities. Matt Botvinick, a former senior director of research at Google DeepMind, will lead work on AI and the rule of law, with plans to build another research team examining how AI will interact with legal systems 3

3

4

4

. Zoë Hitzig, who previously studied AI's social and economic impacts at OpenAI, is joining to connect economics work to model training and development 2

2

4

4

.

Anton Korinek, a University of Virginia economics professor on leave, will lead a project focused on understanding how AI could "reshape the very nature of economic activity" 3

3

4

4

. The institute is also working on projects to predict future AI progress and understand technology's interaction with governance systems.

Policy Advocacy Agenda Takes Shape

Anthropicplans to continue advocating for export controls on advanced chips, a clear federal AI regulation framework, energy ratepayer protections, and model transparency 3

3

. The federal affairs team includes registered lobbyists from both parties, reflecting the company's bipartisan approach to policy engagement 3

3

. Following the Washington D.C. office opening, Anthropic intends to expand its policy work in international markets 4

4

.

The company argues that those building frontier AI have access to information about its risks that nobody else possesses, positioning the institute as uniquely capable of reporting candidly about what it learns 5

5

. Anthropic believes transformative AI isn't decades away but arriving within the next two years, making the institute's work particularly urgent 5

5

. Whether this transparency from inside the industry proves more valuable than external analysis remains to be seen, but the questions the institute tackles—who governs AI advancement, who gets informed when critical thresholds are crossed, how societies absorb AI-driven displacement—demand serious attention as AI safety concerns intensify across the sector.