Responsibility, trust crucial for AI systems: Anthropic India head

Anthropic bets on responsible AI and trust-led development to scale in India, as demand for safer, reliable large language models grows across enterprises

Responsible artificial intelligence (AI) is the right way forward for Anthropic to build its base in India and in today's world trust trumps everything else, the company's India head Irina Ghose said.

 

"I think in today's world trust trumps everything else and that's the way customers want to go in for any conversation," Ghose said in a fireside chat at AWS Bengaluru Summit on Wednesday. 

 

Safety, guardrails, trust in large language models (LLMs) as they hallucinate less, and responsible AI are considered some of the major bedrocks which can improve enterprise AI adoption over time, she added. 

 

"When we think of what we can do taking things across there are two parts: capability and safety. And we treat both of these as the same side of the engineering problem. Nothing comes one after the other and, hence, (both are factored into) what we choose to research on, how we test it, who has an access to it," she said, adding that safety in AI is governed by three principles; honesty, extended thinking and ensuring customer data is safe and not being trained on. 

 

Anthropic, one of the most valuable privately-held AI companies, has committed to spend more than $100 billion over the next 10 years on Amazon's cloud technologies. The e-commerce giant in return will plough in $25 billion into the maker of Claude.

 

Ghose, on her part, has often talked about the importance of trust in LLMs. In February, she said trust is no longer something that one takes for granted. It has to be thoughtfully designed to the core, and not as an afterthought, she said.

 

Anthropic said last year that it was setting up operations in India by opening an office in Bengaluru — its second global office outside of the US after Tokyo, Japan. The company, valued at about $380 billion, will be deploying AI for social impact in sectors such as education, healthcare, and agriculture. Anthropic will also support key industries through strategic partnerships with Indian enterprises, nonprofits, and startups.

 

In six months, Anthropic has infused 10 Indic languages in Claude which include Hindi, Bengali, Marathi and Urdu. Nearly half of Claude usage in India comprises computer and mathematical tasks such as building applications, modernising systems, and shipping production software.

 

Anthropic’s revenue run rate in India has doubled in the last six months and the country is the second largest user for Claude.ai globally.