Anthropic launches the Anthropic Institute, an internal think tank that unifies three existing research teams—Societal Impacts, Red Team, and Economic Research—to study AI’s large-scale societal implications, safety, and economic effects.

The Institute aims to inform policy and safety within the company and will incubate new teams, forecast AI progress, and explore AI’s impact on jobs, governance, values, and our sense of control.

Public policy expansion accompanies the launch, focusing on model safety and transparency, energy ratepayer protections, infrastructure investments, export controls, and democratic AI leadership, with Sarah Heck leading as Head of Public Policy.

Leadership frames the Institute as tackling the hardest questions posed by increasingly powerful AI, anticipating progress toward powerful AI by year-end or early 2027 and addressing societal and ethical impacts.

The announcement references related partnerships and updates, including Mozilla collaboration on Firefox security and other organizational notes.

Institute goals include examining job/economy impacts, societal resilience, AI values, governance, and potential recursive self-improvement, while engaging external audiences to address risks.

Anthropic emphasizes transparency and public benefit, highlighting its status as a public benefit corporation and the belief that safety and long-term research can be a profit center alongside commercial AI development.

The Institute’s debut comes amid tensions with the Pentagon, including a lawsuit over Anthropic’s designation as a supply-chain risk and concerns about DoD restrictions on its technology.

Founding hires include Matt Botvinick on AI and the rule of law, Anton Korinek on economic transformation, and Zoë Hitzig linking economics to model training, with ongoing recruiting for a small analytical staff.

The Institute will publish research for researchers and the public during the transition to more powerful AI systems, addressing major societal challenges from advanced AI.

Described as an internal research arm, the Institute aims to inform policy, safety, and responsible AI development within Anthropic.

About 30 people will join at launch, with plans to study long-term questions on AI’s effects on jobs, safety, values, and human sense of control, and potential for substantial future growth.