Anthropic launches new institute to tackle powerful AI's impact

Anthropic co-founder Jack Clark has a new job title: Head of Public Benefit. Along with the title change, the company just launched the Anthropic Institute, a dedicated initiative designed to study the massive societal, economic, and legal fallout of increasingly powerful AI.

Anthropic isn’t sugarcoating its timeline. The company warns that AI progress is compounding at a breakneck pace, predicting far more dramatic leaps over the next two years than anything we’ve seen in the past five. In their view, the transformative AI described in CEO Dario Amodei’s essay Machines of Loving Grace is landing much faster than the public expects.

Merging red teams and raiding OpenAI

Rather than starting from scratch, the new Institute is consolidating three existing internal research teams at Anthropic.

That includes the Frontier Red Team, which stress-tests models to uncover their outermost limits; the Societal Impacts team, which tracks real-world deployment; and the Economic Research team, which studies disruptions to the labor market. The Institute is also launching fresh initiatives focused on forecasting AI development and analyzing how frontier models collide with the legal system.

To run these efforts, Anthropic made a handful of high-profile founding hires across law, economics, and machine learning.

Matt Botvinick, formerly a senior director of research at Google DeepMind and a Princeton professor, is joining from Yale Law School to lead work on AI and the rule of law. Anton Korinek is taking leave from his economics professorship at the University of Virginia to lead research on how AI reshapes economic activity. Meanwhile, Zoë Hitzig, who previously worked on societal and economic impacts at rival OpenAI, is stepping in to bridge economic insights directly back into model training.

A direct line to Washington

Research is only half the puzzle. Alongside the Institute, Anthropic is aggressively scaling up its Public Policy team.

The policy arm is led by Sarah Heck, Anthropic’s current Head of External Affairs, who previously ran entrepreneurship at Stripe and worked on global diplomacy policy at the White House National Security Council. Her team will tackle hot-button regulatory issues: model safety, energy ratepayer protections, compute infrastructure, export controls, and AI governance.

The real physical signal? Anthropic is opening its first office in Washington, D.C. this spring, planting a direct flag right where tech policy gets written.

Why it matters

On paper, this sounds like standard corporate public relations. But in practice, we think Anthropic is laying the defensive groundwork for what comes next.

For operators and developers, the message between the lines is urgent: Anthropic genuinely believes models with extreme real-world autonomy and labor disruption are 12 to 24 months away. If the Institute actually shares transparent data from its frontier red-teaming, it could give builders a much clearer forecast of what workflows will be upended first.

Politically, Anthropic is playing a clever game. As frontier labs burn immense power and face mounting lawsuits, regulators are circling. By standing up a formal Institute staffed with academic heavyweights alongside a new D.C. lobbying shop, Anthropic is positioning itself as the responsible, data-backed partner to lawmakers.

It looks far more sophisticated than OpenAI’s or Google’s traditional policy posturing. Still, the underlying tension is impossible to ignore: an internal institute funded by a venture-backed frontier AI lab will always face skepticism when its findings conflict with commercial survival.