Anthropic announces the establishment of the Anthropic Institute. This is a new research organization that focuses on the societal implications and risks of increasingly powerful AI systems.
With this initiative, the company aims to share insights about the impact of advanced AI and collaborate with researchers, policymakers, and other parties to better understand potential risks.
The new institute brings together three existing Anthropic research teams. It will be led by co-founder Jack Clark. At the same time, Clark will take on a new role within the company as head of public affairs. According to Anthropic, the institute should become a central hub for research into the economic, social, and security aspects of AI.
One of the teams that will be part of the institute is the Frontier Red Team. This group investigates the security risks of AI systems, including in the field of cybersecurity. In a recent project, the team used the Claude model to analyze the source code of Firefox for vulnerabilities. It then tested whether the AI could independently develop methods to exploit such vulnerabilities.
The Societal Impacts team will also become part of the institute. This group analyzes how users deploy AI systems in practice. The team recently published research on situations in which employees choose to let AI agents perform tasks completely autonomously, without direct human control.
The third team within the new institute is Economic Research. This team investigates the economic impact of AI and publishes, among other things, Anthropic’s Economic Index. This report analyzes which business activities customers automate using Claude and how AI is applied in different sectors.
Anthropic attracts prominent researchers
In addition to bringing the three teams together, Anthropic wants to further expand the institute with new researchers. Several prominent names have already been attracted. Matt Botvinick, former research director at Google DeepMind and professor at Princeton, will lead research into the relationship between AI and the rule of law. Economist Anton Korinek, from the University of Virginia, will focus on research into how transformative AI can change the nature of economic activity. Zoë Hitzig, who previously worked at OpenAI on research into the social and economic impact of AI, will help to link economic research more closely to the development and training of AI models.
According to Anthropic, the institute is also working on projects that attempt to better predict future AI developments. It is also investigating how advanced AI systems could interact with legal systems in the future.
Parallel to the launch of the Anthropic Institute, Anthropic is also expanding its public policy organization. This team focuses on policy issues related to AI, including model safety, transparency, infrastructure investments, and export regulations.
The policy department will be headed by Sarah Heck, who was previously responsible for entrepreneurship policy at Stripe and was involved in economic and diplomatic policy within the US National Security Council. The company plans to open its first office in Washington, D.C., this spring to position its policy work closer to US decision-makers. Anthropic then intends to further expand its policy activities internationally.