Former Y Combinator president Geoff Ralston launches new AI ‘safety’ fund

https://techcrunch.com/feed/ Hits: 25
Summary

Geoff Ralston, well-known in the startup community for his years at Y Combinator, is back in the formal investing ring, he announced Thursday. His new fund is called Safe Artificial Intelligence Fund, or SAIF, which is both an explanation of its thesis and a play on words. Ralston is specifically looking for startups that “enhance AI safety, security, and responsible deployment,” as his fund’s website describes. He plans to write $100,000 checks as a SAFE, “pun intended,” he says, with a $10 million cap. A SAFE is, of course, the invest now/price later pre-seed investment tool pioneered by Y Combinator (it stands for simple agreement for future equity). While most VCs these days are looking to invest in AI startups, Ralston’s take is a bit more focused on the idea of safe AI, even though he admits the concept is a bit broad. “The vast majority of AI projects out in the world today are using the technology to solve problems or create efficiencies or create new capabilities. They are not necessarily intrinsically unsafe, but safety is not their primary concern,” Ralston tells TechCrunch. “I intend to fund startups whose primary objective is safe AI — as I have (very broadly) defined it.” That list includes startups focused on improving the safety of AI, like those that clarify an AI’s decision-making process or benchmark AI safety. It includes products that protect intellectual property, those that ensure an AI conforms to compliance requirements, fight disinformation, and detect AI-generated attacks. He also wants to invest in functional AI tools with built-in safety in mind, such as better AI forecasting tools and AI-enabled business negotiation tools that won’t reveal corporate secrets to outsiders. This might sound like a list of AI startups that many VCs are pursuing, but there are areas Ralston says he won’t back. One example is fully autonomous weapons. “There are certainly uses of AI which would (will) be unsafe: using the technology to create bioweapons, to m...

First seen: 2025-04-17 20:14

Last seen: 2025-04-18 20:18