That's not really comparable; by reverse-engineering a cookie one would have to fight against entropy. The issue with software is that it's magnitudes easier to

That's not really comparable; by reverse-engineering a cookie one would have to fight against entropy. The issue with software is that it's magnitudes easier to figure out any trade secrets inside of it. Reverse-engineering LLMs is much more akin to the cookie example.