I'm in the insurance business -- an expensive place to be in the past couple of months. It was made costly for Berkshire Hathaway, the company I run, because I did something very dumb: allowed Berkshire to provide insurance coverage for a huge catastrophe loss without its getting a premium for doing so. The risk we unthinkingly assumed was a loss from terrorism.

Given the degree of my error, I was lucky: We estimate our Sept. 11 loss at Berkshire to have been "only" about $2.3 billion. That's more, by far, than we've ever lost from a single catastrophe, but the toll could have been far larger. Indeed, had a nuclear device been available to Osama bin Laden, the loss could have bankrupted most of the insurance industry, Berkshire very much included. Given that kind of horrendous, but not impossible, nuclear scenario, insured losses could have been $1 trillion, an amount that exceeds the net worth of all property-casualty insurers worldwide.

A potential loss of almost infinite magnitude can be assumed only by an entity of almost-infinite resources. That economic species doesn't exist in the private sector. Only the U.S. government fits the bill.

Washington has accepted this proposition to a point. Congress is now agonizingly trying to create some sort of industry-government coalition that would insure losses from terrorism. The motives of all concerned are admirable, but I believe the actions now being considered have disturbing implications for our society.

Some proposals limit the government's liability but leave the risk for the industry open-ended. These proposals won't work: If unlimited liability is left with insurers, they will necessarily refuse to renew policies they see potentially leading them to bankruptcy.

Equally bad, all of the proposals now being considered will engender pricing based upon risk exposure. That is, if a business is located in a high-risk spot it will be asked to pay a staggering price for insurance. Risk-based pricing is normally equitable and desirable. At Berkshire, we employ it constantly. In this case, though, it would have antisocial consequences.

For example, the terrorism risk per dollar of insured value may be 10 or more times for iconic or critical properties in New York City what it is for properties in less-populated areas. But great cities are central to our society. We don't want them to wither under the burden of hugely disadvantageous insurance costs. Indeed, it's in America's interest for them to thrive. Citizens of our leading cities almost certainly bear above-normal physical risks in the war being waged upon us by terrorists. We should not impose crippling economic costs on them as well.

In my opinion, we would arrive at a solution for this societal problem if we were to adopt the Federal Deposit Insurance Corp. as a model for where we want to head in the insurance industry. The rationale for the FDIC, formed 68 years ago, was clear-cut: The United States sorely needed to eliminate bank runs and the financial panics they caused. Prior to the FDIC, the risk from bank failures resided with depositors, who had no way to shed it. Neither they nor their banks could lay that risk off on private insurers for two reasons: First, the dollar amounts involved were simply too large; second, losses were correlated, in the sense that the failure of a few banks frequently caused a chain reaction, in which good banks toppled with bad, leaving a mountain of economic damage. Fortunately for the country, these punishing disruptions to our economy were ended by the advent of FDIC insurance.

At the moment, leaving aside insurance policies soon due to run out, millions of business owners, individuals, landlords and lenders bear the economic risk of terrorist attacks. Insurers won't step up to assume the risk -- we were previously dumb, but we've learned. It isn't right, though, that these risk-laden millions should have to shoulder this burden themselves: That would be self-insurance, and the economic distortions it would cause would stagger our society. Who would ever build a skyscraper in a major city or lend against it? Or how could a mom-and-pop deli, bare of insurance, locate next to an iconic office tower?

If we were to adopt an FDIC model for handling terrorism, the insurance industry would not be permitted to earn a dime from the coverage. Instead, a premium tax, payable to the U.S. Treasury, would be levied on all insurance. This would have the equitable effect of spreading the terrorism-related cost to the country in general, just as we spread defense expenditures.

Were a proposal such as I suggest to be enacted, the new law should sharply limit private lawsuits seeking to place blame on some party involved -- an airline, say. We should want the Treasury to make payments to victims solely to compensate them for loss of property, life or direct earnings, without worrying about fault. The law also should cover war losses. War and terrorism both have the capacity to impose losses that private insurers simply can't bear. Most existing policies already have a war exclusion -- but not all by far.

Some people will argue that an FDIC model for insurance would be a socialistic intrusion into the private sector. Most surely would have argued the same about the FDIC itself -- and yet that institution is today generally regarded as having been enormously beneficial. The problem was once bank runs and economic panics, and we found an innovative solution. Today, the problem is terrorism and its capricious effects on insurance costs, and we need a solution of comparable efficacy.

The writer is chairman of the board of Berkshire Hathaway Inc., a diversified company with insurance operations, and a director of the The Washington Post Co., which has an investment in Berkshire Hathaway.