This AI Finds Patent 'White Space' — Unexplored Inventions Waiting to Be Made

Every big company wants to invent something new — but they also want to avoid trespassing on someone else's patent. That's where 'patent white space' comes in: it's the empty territory between existing patents, where a company could build something novel without getting sued. Traditionally, finding that space meant hiring experts to manually review thousands of patent documents — slow, expensive, and easy to miss things.

BLANC, a new AI method from Japanese researchers, automates this search. It reads patents through three different lenses: what the invention is used for, how original it is, and what makes it non-obvious. Then it measures how strongly those dimensions connect. When a combination of features is common globally but missing in a specific area — say, 'fluorine' and 'warpage suppression' in glass production — BLANC flags it as a promising white space.

The tool proved itself in two ways. First, on public patent databases, researchers deliberately removed chunks of known invention pairs to see if BLANC could find them again. It recovered roughly 34% of the hidden combinations in AI patents and 27% in glass patents — while random removal never uncovered anything. Second, in a real industrial case, BLANC flagged a specific glass treatment idea that human experts had independently identified. That's a strong signal the tool finds real opportunities.

For everyday people, this is more than a back-office trick. Faster, cheaper discovery of patent white space could mean more new materials, medicines, electronics, and consumer products reaching the market sooner. It also means companies can spend R&D money on genuinely new territory instead of re-inventing what already exists — potentially lowering costs and accelerating innovation.