A startup called Probably has raised $9 million to tackle one of artificial intelligence's most stubborn problems: the tendency of AI systems to confidently state things that are simply not true.
According to TechCrunch, Probably wants to prevent hallucinations and factual errors from ever reaching users — aiming for a level of accuracy that matches so-called deterministic systems, which are traditional software programs designed to produce the same correct output every time they run.
The funding signals growing investor appetite for solutions to AI reliability, a challenge that has dogged the industry since large language models went mainstream. While AI tools have proven useful for drafting text, answering questions, and automating tasks, their habit of inventing facts — often with complete confidence — has made businesses and professionals hesitant to rely on them for anything high-stakes.
Probably's ambition is significant: closing the gap between the fluid, probabilistic nature of modern AI and the hard accuracy guarantees people expect from conventional software is a problem researchers have wrestled with for years without a clean solution.
If Probably can deliver on that promise, it could unlock AI adoption in fields like healthcare, law, and finance, where a single fabricated fact can have serious consequences.