But if the drills end up killing thousands or the pen tests end up causing massive data breaches, wouldn't you agree that the pen tester or drill organizer should be held liable? Or do you just let them do it again because it's just a risk we need to take

If you leak pathogen - sure. But there’s no proof of any kind that it happened, and people fear mongering greatly overestimate abilities of gain of function. It’s like Skynet vs current state of an AI.

We don’t ban pen tests, just because people who have 0 understanding of it are sacred of it.

There have been many documented instances of lab leaks. It's even happened with Covid-19 research[1] at least once that was reported, and with the virus being pandemic around most of the world I'd assume there were many other leaks that weren't reported because nobody could prove it came from a lab. And even if the evidence that the original Covid-19 resulted from a lab leak is weak the evidence that the Omicron variant came from a lab is much stronger.

Scientists are only human and bio research labs don't have the same culture of continuous improvement and lack of blame that have let the airline industry, say, get to its current level of safety.

EDIT: Oops, forgot to add the citation

[1] https://fortune.com/2021/12/10/taiwan-investigates-covid-lab...

There is obviously a risk of pathogen leak..

https://armscontrolcenter.org/wp-content/uploads/2016/02/Esc...

And there are highly qualified experts in the field who view GoF research as the most likely source of COVID-19, so on what basis are you so confident that it doesn't pose a major risk?

Lab leaks (even without GoF involved) are remarkably common events.

So shouldn’t be base our regulation on that reality?

gain of function research is more like inventing novel nuclear weapons more likely to kill humans to study bomb shelter and anti radiagion technology than any drill. the pathogens created are not drills.

Only if nuclear weapons would evolve by its own, and on insanely much higher speed than what you can invent.