The CEO of Perplexity said the AI-powered search engine isn’t trying to replace the news

Perplexity CEO Aravind Srinivas.

The simmering tensions between news publishers and search engines built with artificial intelligence spilled into public view Wednesday on a beachside conference stage.

“Let’s talk about the elephant in the room,” Aravind Srinivas, cofounder and CEO of Perplexity, opened his interview at Wall Street Journal’s Tech Live conference.

Rupert Murdoch’s Dow Jones, publisher of the Journal, and the New York Post filed a lawsuit in New York on Monday accusing Perplexity of “freeriding” on the journalism they produce. Perplexity’s app allows users to get instant answers to questions with sources and citations from trusted blogs, news outlets, and academic papers.

The two-year-old startup takes in cash like oxygen, having raised back-to-back funding rounds just months apart, and is said to be in talks to raise a fourth round that would value the startup at $8 billion.

The lawsuit alleges that Perplexity is scraping copyrighted work to feed its search engine and, in doing so, is routing would-be readers and customers away from their websites.

“Nobody’s coming to Perplexity to ingest their news. People are directly going to [The] New York Times, The Wall Street Journal,” Srinivas said. He explained Perplexity’s goal as helping users digest the news. “People come to Perplexity for understanding the news in the context of what they already know. ‘How does this news affect me?'”

“We are not interested in taking the exact content and resurfacing it. We are not trying to be the news alternative,” he later added. “So, we are going to try our best to engage and communicate what the goals of our product are and how they can be symbiotic with the existing news outlets.”

The company has said it does not scrape data for training large language models; rather, it scours the internet, creating an index of web pages for its models to reference.

The entrepreneur stayed coolheaded as WSJ reporter Deepa Seetharaman lobbed questions about Perplexity’s cash flow and claims by news outlets that it doesn’t make citations clear. Srinivas acknowledged the tech is “not perfect.”

“We are still working with a piece of technology that is constantly improving,” he responded to a question about plagiarism. “Even yesterday, Anthropic released another version of their models, Claude 3.5. It’s getting better and better every few months. So whatever issues exist today are a new set of issues that did not exist one or two years ago.”

In his interview, Srinivas defended the company’s modus operandi, saying it had a conversation with the Journal’s publisher in June about potentially striking up a contract that would allow Perplexity to share future ad revenues with the publisher. He didn’t specify the contract’s contents but said the conversation went dark.

“We certainly were very surprised about the lawsuit because we actually wanted a conversation,” Srinivas said.

This month, the company will start selling ads against popular searches, Srinivas said. So, for example, a user can ask Perplexity about a shoe company that has just gone public on Wall Street. The search engine will return a summary of the shoemaker’s stock debut with sources and citations. If a shoe brand pays to advertise in the results, Perplexity will share a cut of the money it makes with publishers whose content it used to answer the query.

“We can only exist if we make money of our own, too, not just keep fundraising,” Srinivas said. “Then when we start making money of our own, that will certainly benefit all the publishers and ensure they also continue to flourish.”

Just last week, The New York Times sent Perplexity a cease-and-desist letter seeking to end the practice of using its stories to train chatbots. The news outlet is also suing OpenAI and Microsoft for copyright infringement.

Perplexity hopes to turn profitable in the span of three to five years from selling subscriptions and advertisements, according to Srinivas. He acknowledged that while Perplexity isn’t training large language models and doesn’t need to raise as much capital as an OpenAI or Anthropic, the influx of investor cash allows it to move more quickly.

“Fundraising obviously helps us to stay focused on the product and keep improving the product but not be overly worried about the financials,” Srinivas said.