Is Any Infringement a Good Infringement?

Is Any Infringement Good Infringement?

In some recent benchmarking, I’ve seen noticeably less convergence among AI-based infringement detection tools than in prior tests. I’ll have more to report on that later, but it prompted me to step back and revisit a question that feels central to this topic.

When we’re looking for infringement, what exactly are we optimizing for? Are we trying to surface infringement by larger entities with meaningful revenue streams and a real royalty base? Or are we just as interested in identifying smaller, sometimes almost incidental, infringements that may be technically correct but commercially marginal?

I’ve had a clear philosophy on this for a long time, but it’s worth revisiting, especially in the context of AI-based infringement detection tools, which can be implicitly “calibrated” toward finding either any infringement or only significant infringement.

In my experience, the reality is that relatively few patents are infringed, and even when they are, infringement is not always easy to detect and document. For example, in an industrial portfolio coming out of a corporate R&D effort, it wouldn’t be unusual for only 8–10 patents out of 100 to suggest infringement, and after researching products and developing claim charts, to have no more than 5 of those patents actually read on products. One would be very lucky if say, 5 of the patents read on the products of 3 companies, resulting in 15 solid claim charts.

Given those statistics, I’ve long taken the view that any infringement is a good infringement; at least as a signal. It tells you the claimed technology is actually being used in the real world.

Even when the associated revenues are small and the business case for licensing or enforcement isn’t there, I don’t view it as noise. It’s data. It shows how the technology is being implemented, where it’s showing up, and how it’s finding its way into the market.

In that sense, dismissing “small” infringements risks overlooking some of the clearest evidence that a patent is doing what it was supposed to do.

The other side of the coin, however, is the view that if an infringement is too small, it’s simply not worth the effort. Under that approach, the focus is on specific instances where infringement can not only be proven, but where a compelling business case immediately follows, namely, a substantial and growing royalty base.

With AI-based patent infringement detection tools, this calibration is becoming more pronounced. Some tools are effectively tuned to surface any infringement, while others prioritize larger, higher-value targets.

Relying on a single tool can skew the picture, since you’re implicitly buying into that tool’s underlying methodology. Which raises the obvious question: do we need to be using multiple tools to properly assess a portfolio?

We’ll see shortly.

In the meantime, I’d love to hear your thoughts on this, so please comment or reach out to me directly.