Well, riskiest experiment to the largest number of people ever? In terms of what we know about, that would be the work that lead to the atom bomb. You can pretty much assume that there’s probably far scarier stuff we don’t know about that’s been/is being mucked with, but it’s a little hard to comment on that, obviously.

You could also argue that something like Colombus’ voyage was a bit of an experiment that ended in disaster for quite a few people.

The atom bomb is a little unexciting though, and everyone knows about it, and Colombus is a little tenuous at best in regards to the type of answer you seem to be looking for. So I’d have to go with something like “gain of function” research.

It’s not a specific experiment so much as a field of science – in this case, virology.

The general, likely incredibly over-simplified version is this: somewhere out there are scientists working tirelessly to mutate viruses into the most possibly dangerous versions of themselves possible.

The idea is that if we can make things far more lethal than they’ve become in nature thus far, we’ll have a better understanding of how they work and how best to treat them should they eventually mutate along similar lines via natural methods. The end goal, according to many researchers, is to better develop broad-spectrum anti-virals and medication – at present, most vaccines, for example, are fairly specific to a couple of strains of a virus, but are unable to do much against other strains of the same thing. While this matters very little at the moment, some of these other harmless strains could someday pose a pretty big risk.

So, somewhere out there, scientists are doing things like using the swine flu to recreate a virus that killed more people than the second biggest war in history, or making the bird flu both airborne far more virulent in humans despite the fact experts have stated a more virulent version of it could kill half the human population.

“I can see the potential benefits,” you say, “And surely, we have the best and brightest working on this in secure, remote locations. I get that this *could* be scary, but c’mon.”

Well, here’s what your average “secure location” looks like:

As you can see, it has doors that have locks on them, and we can assume an alarm system probably exists. You’ll notice though, that it looks distinctly less bunker-y than I think most of us picture these sorts of things – no gates, or fences, or armed guards, etc.

It’s not particularly remote, either:

You’ll notice a McDonald’s about a block away, for example.

This lovely building also happens to have been the site of a series of “oops, we let it out” type events. Fortunately for rich old condo-dwellers and big mac enthusiasts alike, the “it” in this case, a strain of ebola that is incredibly transmissible in primates, turned out to be non-dangerous to humans. Perhaps somewhat un-intuitively, after the research center was shut down, it was converted into a daycare center – fortunately, we know for an absolute fact that stuff like samples of incredibly deadly viruses getting left behind and forgotten about for some kid to find never happens.

Most of these sorts of centers are in, or near, major population centers, much like the Reston facility above. And while you’d think something like the above happening, especially in top notch federal labs staffed by some of the best minds the west has to offer, would be a once-in-a-lifetime event, it turns out that even despite best efforts, training, and diligence of the people doing this kind of work, mistakes actually happen all the time. Even when they don’t, some of the viruses used in this sort of work are so dangerous that even with proper procedure and no mistakes, they have a 10–20% chance to escape and “spread widely” just via regular handling.

So, that’s my thought. It’s not just scary because it can certainly kill us all – which, unlike the LHC or other “scary” experiments, is a 100% known potential outcome. It’s scary because if it does it, it will likely be accidental and because someone did something minor like forget to put some tape on a seam.