Anthropic launches institute to study AI's societal risks

Anthropic has launched the Anthropic Institute, a new research and policy-facing group that will publish work on the societal challenges posed by increasingly powerful AI systems and expand engagement with external audiences.

The initiative brings together several existing research teams and adds new programmes focused on forecasting AI progress and analysing how advanced systems may interact with the legal system. The Institute sits alongside an expanded Public Policy organisation, as Anthropic argues that AI development is accelerating and that significant gains in capability could arrive soon.

Anthropic, a US-based AI company known for building large language models, expects "far more dramatic progress" in the near term. It pointed to recent advances, such as identifying severe cybersecurity vulnerabilities and performing a wider range of workplace tasks.

Mandate And Scope

The Anthropic Institute will draw on research from across the company and share findings with researchers and the public. Its focus spans labour markets, economic change, societal resilience, security risks, and questions about the "values" AI systems express and how companies set them.

It also plans to address governance issues that could arise if AI systems begin to improve themselves recursively. The topic raises questions about when such developments should be disclosed and what oversight should apply.

Anthropic framed the Institute as a means to enable more candid reporting from organisations developing frontier AI models. Developers of leading systems hold information that outside groups do not, including details from safety testing, deployment experience, and model behaviour under stress.

Leadership And Teams

Jack Clark, an Anthropic co-founder, will lead the Institute and serve as Head of Public Benefit. The group has an interdisciplinary staff that includes machine learning engineers, economists, and social scientists.

The Institute consolidates and expands three research teams. Frontier Red Team stress-tests AI systems and examines their limits. Societal Impacts studies how AI is used in the real world. Economic Research tracks effects on jobs and the wider economy.

It will also incubate new teams, including efforts to forecast AI progress and analyse how powerful AI systems may interact with legal processes.

Founding Hires

Anthropic has made several founding hires for the Institute. Matt Botvinick, a Resident Fellow at Yale Law School and former Senior Director of Research at Google DeepMind, will lead work on AI and the rule of law. He was previously Professor of Neural Computation at Princeton.

Anton Korinek will join the Economic Research team while on leave from his role as Professor of Economics at the University of Virginia. He will lead an effort studying how "transformative AI could reshape the very nature of economic activity," according to Anthropic.

Zoë Hitzig, who previously studied AI's social and economic impacts at OpenAI, is joining to connect economics research to model training and development. Anthropic is also hiring a small analytical staff to synthesise parts of its research agenda and communicate results.

External Engagement

Alongside publishing research, the Institute plans to engage more deeply with groups that expect disruption from AI adoption. This will include dialogue with workers and industries facing displacement, as well as communities anticipating rapid change.

Anthropic said the exchange will shape the Institute's agenda and broader corporate decisions. The two-way approach reflects a growing trend among AI developers to pair technical safety work with social research and stakeholder engagement.

Policy Expansion

Anthropic is expanding its Public Policy organisation to address areas where it has "priorities and perspectives". These include model safety and transparency, energy ratepayer protections, infrastructure investments, export controls, and what it calls democratic leadership in AI.

Sarah Heck, currently Head of External Affairs at Anthropic, will lead the function as Head of Public Policy. Before joining Anthropic, she was Head of Entrepreneurship at Stripe and previously led global entrepreneurship and public diplomacy policy at the White House National Security Council.

Anthropic will open its first office in Washington, DC, in the spring and is expanding its global policy footprint.