It's mostly good, though there was some bad/ambiguous wording in the original AIA that is now "mostly fixed".
The thing to remember is that first-to-file doesn't mean that if i file a patent app on something you invented, before you do, that I get a patent.
It's about what happens when two people file patent apps for things they can both prove they invented.
Before, there was a complex system of colliding patent resolution. Now, it's "first guy to file wins".
More things are also now prior art. You can't get around prior art by showing you invented before the prior art was published, for example.
> The thing to remember is that first-to-file doesn't mean that if i file a patent app on something you invented, before you do, that I get a patent. It's about what happens when two people file patent apps for things they can both prove they invented.
was very helpful in clarifying.
Another failure mode is that computer spies steal the IP via a software security flaw, then patent it as their own independent reinvention.
Let's say I invent a new search engine, Poople. I don't have much money to patent it. But I can't keep the details secret either lest somebody else patent it. But by publishing the details defensively, my competitor with deep pockets, Paapo, goes and reimplements my algorithm and wipes me out of the marketplace.
Which part of the story did I get wrong?
Commercial sales by competitors are not subject to patent during that interim period, and they can continue such sales after the patent application is filed, but once the patent application is filed they can only continue to sell without a license the exact same products they sold before the application. So if the competitor changes the product sold, they could be subject to the patent (assuming the patent is granted).
As for your specific example: algorithms can't be patented; they are trade secrets. Specific software implementations of algorithms possibly can be patented. The difficulty, of course, is that deep pockets Paapo can always engineer around the specific software implementation you specify in your application.
That standard has not changed at all.
Also, you're suggesting that if two criminal activities were well conducted, the inventor will lose a civil case? If this criminal activity is revealed, it invalidates the patent and subjects the executives and lawyers responsible to criminal sanctions and massive civil liabilities, including forfeiting all profits derived from the fraudulent patent and treble damages (i.e., 3x the actual damages the true inventor suffered). It's simply not ever worth the business risk to pursue this path.