It seems to me that an individual/corporation can present the world with significantly more liability than they could ever repay. I was curious how this is or could be prevented?

For example, (don't get hung up on details, just general concept).

You have some company that is collecting/producing a hazardous material. The company is hypothetically worth a billion dollars or something. Then lets say that there is a small chance they could have some huge accident that would cost 500 Billion to society through deaths/cleanup/property loss/cancer, etc.

In this case even if all the assets of the company went to repayment, society would still be in the hole billions of dollars. Is there a way to prevent this corporation from presenting risk that it would never actually be able to fix?

No; they will just buy off the appropriate politicians and bureaucrats.

There are lots of possibilities: Fight them, do business with them anyway, refuse to do business with them, tell your friends not to do business with them, provide the same service in a safer way.

Different strategies are appropriate in different situations.

This is already called "judgement proof', and it happens to individuals as well as corperations.

This already happens, particularly in corporations that go into third world countries but have limited assets in those countries. After some humanitarian or environmental disaster, IF the country and victims have the legal capacities to actually get a guilty verdict (a big if) then by that point the remaining assets in the country are dwarfed by the costs. While corporations often have more than enough assets outside this particular country it is very hard in international law to go after this.

Insurance and re-insurance.

This can turn out like the credit derivative market in 2008 where the insurer basically becomes the company described in the OP, i.e. $x of liability exposure but a much smaller market cap.

OP, how much liability is too much? How likely is too likely? Even if you can decide on this, who's going to build a risk profile to determine if the risks are too great? Modeling "tail" risk is extremely difficult.