San Francisco-based Harvey AI, the world’s most valuable legal artificial intelligence (AI) startup, has set up full-scale engineering and policy operations in India, as demand for the technology grows among the country’s top law firms.
India ranks among the legal-tech startup’s top markets around the world, chief technology officer Siva Gurumurthy said in an interview. To cater to the demand, the company has built an agentic AI platform where law firms and clients can collaborate on ongoing legal operations, litigation and policy requirements.
“We have a technical team, as well as a full-scale product development team and a go-to market team in India, which work with various law firms, local partners and corporate clients to fine-tune our models and platforms for what companies require in the country. Each country’s legislative system is different from another’s; so, we’ve refined and tuned our models to have knowledge relevant to each market around the world,” Gurumurthy said.
Founded in 2022, Harvey AI received early backing from OpenAI’s startups fund in the US. The company incorporated its India entity on 8 November last year. Its early clients include law firms Shardul Amarchand Mangaldas, AZB & Partners and PricewaterhouseCoopers (PwC) India.
Lawyers deploying AI into their workflows clearly see its value. Vijayendra Pratap Singh, senior partner and head of dispute resolution at AZB & Partners, said that while the potential of AI platforms in law firms is expansive, a human-in-the-loop approach is key to maximizing productivity while retaining the nuance and accuracy of results i.e. meaningful accountability.
“There is a clear impact that customized AI solutions can and will have on the field of law. However, lawyers need to ensure trustworthiness of AI solutions. In the Indian context, the biggest factor to take into consideration is that India simply does not approach courts unless absolutely necessary i.e. litigation is a last resort. AI cannot come up with the right strategic implementations of laws and legal arguments in cases—lawyers, on this note, will always remain the ones leveraging it. An AI platform will only help in improving process efficiencies, while how a platform is implemented in high-sensitivity cases will be reserved for the ones with the experience to deploy effective strategies,” he said.
On 25 March, Harvey AI raised $200 million at a valuation of $11 billion, making it the most valued company building AI models and platforms for use in law firms and courtrooms. Harvey is backed by Sequoia Capital, Singapore’s sovereign fund GIC and Andreessen Horowitz, among others.
“We also have a legal agentic AI platform that works like a review table between law firms and their clients. This allows a corporate entity to collaborate with their law firms and legal partners in real time, in terms of reviewing documents, legal arguments, or particulars of a case—in effect, saving companies hundreds of hours of manpower effort. In India, we’re already working with clients such as Flipkart and the Reliance group, among others,” Gurumurthy said.
“We also have at least three or four different products that we’re building out of India right now,” the executive added.
Since the advent of generative AI among mainstream businesses, legal AI assistants have been growing in prominence. This has led to the creation of homegrown legal AI startups such as Lucio, as well as global ones such as Sweden-based Legora and Harvey itself.
Confidence in AI use among law firms appears to be on the rise. In May last year, Mint reported that law firms were increasingly wary of using generative AI due to concerns around hallucination of the results that AI models were producing. However, many law firms have continued to experiment with AI adoption—in June and September last year, law firms SAM and AZB announced their collaborations with Harvey.
For now, though, Gurumurthy said AI adoption in the legal services industry may not involve the build-out of language-specific AI models, or if courtrooms will adopt AI services themselves. “I can’t tell if courtrooms will adopt AI platforms for their own usages—this is too far into the future.”
While Harvey’s current AI models can understand languages beyond English and ingest them into their databases, he said language-specific capabilities have yet to be built. “Our platform is pretty language-agnostic, and we do localize our platforms for various markets. In future, we’ll have local language models work in various local language courts, but for now, we deploy a set of language-agnostic models in our platform that differ by capability.”