Anthropic flags risks of autonomous weapons, mass surveillance: Amodei

Anthropic has raised concerns about the development of fully autonomous weapons without a human in the loop and the mass surveillance of American citizens, its co-founder and chief executive Dario Amodei said.

Speaking to a television channel on the sidelines of the India AI Impact Summit on Thursday, he said these red lines are critical and that it is essential to ensure AI is used in a way that is compatible with democracy and aligned with the company’s values.

He was responding to a question on whether the company could face blacklisting by the US government because of its restrictions on how the technology is used. “We continue to have productive discussions and are excited to help countries in national security,” he said in the interview.

Addressing the sell-off in IT stocks amid fears around AI models, Amodei said the company is not trying to replace the existing industry but to work with it. He compared Anthropic to Lego blocks, saying other companies build on its models. However, he acknowledged that some disruption is inevitable.

Earlier, speaking at the summit alongside Nandan Nilekani, co-founder and chairman of Infosys, Amodei said India could see 20–25% economic growth driven by AI-led productivity gains.

“It sounds absurd, and is unknown anywhere in the world. But as I think about this, it kind of stacks all the factors for a very bullish picture of how that growth could happen. India has technical capabilities and eagerness to adopt AI. Potential economic growth could be much higher than that of developed countries, and it can do a lot of catch-up growth. India actually almost seems like a perfect case study that AI can accelerate economic growth because AI will help tie together the base ingredients already present here,” he said.

He added that Anthropic is making a concerted push to improve its models’ performance in Indic languages, with a focus on including speakers of less common languages. The company has collated large datasets of Indic languages in partnership with local organisations, and that effort is reflected in the performance of its model Sonnet 4.6.

Anthropic is also partnering with Indian non-profits and initiatives such as EkStep and OpenAgriNet to deploy Claude for rural, farmer-focused and public-good projects, aiming to scale inclusive AI experiments across the country.