Whole life insurance payouts occur only if the insured dies, on a date that is not known to the investor in advance. The value of a bet depends on its date estimate, which is now increasingly done by machines.
getty
There are corners in the market where investors make money when strangers die, and the better you guess the timing, the more profit you make. This transaction is called a life settlement. Investors buy life insurance for seniors, inherit the premiums, and receive a death benefit when the person dies. The entire return is based on one prediction: how long the insured will retire. For most of the market’s history, doctors and actuaries made such decisions. That effort is now starting to move toward machine learning models, and that’s what this article will cover.
take out a stranger’s life insurance
Regulators who investigated this market didn’t hesitate to say how unsettling it was to people. The Securities and Exchange Commission’s 2010 staff report, still the most thorough official report, quotes investors who admitted discomfort with “assets whose gains are tied to death.” But the logic is simple. An older policyholder who no longer wants coverage can sell the coverage for a higher price than the insurance company would pay to surrender, and the buyer ultimately owns the claim for the final payment. The Supreme Court welcomed this practice, dating back to 1911, when it ruled that life insurance policies are treated as property and can be sold like other insurance.
Supply has increased enormously. As America’s baby boomer generation ages, billions of dollars in life insurance policies are canceled or lapsed each year, often at a fraction of the amount traded on the secondary market. The pitch to investors is that returns, often quoted in the high single digits to low teens, have nothing to do with stock markets or interest rates and everything to do with mortality rates, making them an attractive diversifier when stocks and bonds move together. Policies are bundled, bond-financed, and sold primarily to institutions and wealthy individuals, but sometimes ordinary savers are also drawn in.
The modern secondary life insurance market grew in the wake of the AIDS crisis in the 1980s, with terminally ill patients selling insurance to raise money for treatment, and has since expanded to include older people who simply no longer need coverage. The SEC report notes that by 2007, near its peak, about $12 billion in face value of life insurance was being settled per year, before the financial crisis and banks such as Goldman Sachs and Deutsche Bank exiting the business slowed the exit from the life insurance business. These pools have been traded for years by hedge funds and offshore vehicles, sometimes through products built on longevity indices, and even market advocates acknowledge the discomfort of assets paid out on death.
The numbers it all depends on
One of the numbers that determines whether an investor wins or loses is the life expectancy of the insured. The SEC clearly stated so. Estimated life expectancy “constitutes an important component of life insurance contracts, influencing the amount paid to policyholders, the expected timing of payments to investors, and the value of securitizations,” the report said. When you buy insurance for someone who expects to live four more years, the math works one way. Learn that they may live to be nine years old and lose out on the same insurance.
The reason is premium. The investor in one of these policies pays the insurance company perhaps 5% to 10% of the insured amount to keep it in force and gets nothing back until the insured dies. The SEC report notes that it could take more than three years for death benefits to be paid. The longer you wait, the higher your premiums will be, the less profit you’ll make in the end, and eventually the insurance that seemed like a good deal will gradually run out. Simply put, the effect of this transaction is “waste of assets.” For investors, a long life span is a bad outcome.
Machines now guess
The people who make that estimate are called life expectancy underwriters, and the entire deal depends on them. They read insureds’ medical records and return numbers, but the SEC noted significant gaps in oversight. Unlike the brokers and providers around them, “life expectancy insurance companies are not subject to significant regulation at the state level.” The least viewed link in the chain determines the price.
That connection is now being rebuilt around machine learning, and regulators are taking notice. In a model bulletin adopted in December 2023, the National Association of Insurance Commissioners told insurers that AI technology is now being deployed across the insurance lifecycle, including “underwriting and pricing,” and defined the predictive models in question as mining historical data “using algorithms and machine learning to identify patterns and predict outcomes.” The Society of Actuaries, the profession’s own research group, reported in June 2025 that predictive analytics is being applied to “mortality, longevity, and mortality risk models,” the very machines that generate life expectancy.
It’s worth keeping that distinction in mind, as these documents broadly discuss life insurance underwriting, rather than being specific to the life insurance niche. Taken together, they show that the task of estimating a person’s life expectancy, on which the entire value of a community depends, is moving away from actuarial tables and toward machine learning models trained on medical records, prescription histories, and other data. Some longevity insurance companies go further and sell AI-driven “aging clocks” that read biological markers rather than date of birth. Promises are sharper guesses, and sharper guesses truly help everyone in the chain. An investor’s exposure does not change in type, only in source. The value of assets still depends on predictions, which are increasingly generated by systems whose workings are opaque even to those who rely on them.
This model has moved money before
The market has already experienced a version of this. The SEC report notes that revisions to the methodology used by major life expectancy underwriters have historically resulted in lower settlement amounts purchased prior to the change. As human life expectancy estimates become longer, any insurance that was priced based on older, shorter estimates will suddenly drop in value, and investors who buy in good faith will suffer losses. No one died on the other schedule. The model simply changed its mind about the schedule.
The transition to AI underwriting will likely result in new, and potentially larger, revisions. If a new generation model systematically reads longevity differently than the actuarial tables it replaces, portfolios priced based on old assumptions may be increased or decreased accordingly. For an asset sold for stability and distance from market noise, few buyers are paying attention to the volatility it causes.
When it reaches the hands of general investors
The clearest warning of what will happen if this product is made available to ordinary retailers has been laid out in federal court. GWG Holdings, a Dallas company, raised about $1.6 billion by selling L-Bonds, high-yield bonds backed primarily by life insurance proceeds, to about 26,000 investors. Many investors are retirees or conservative savers with incomes. The company filed for bankruptcy in 2022. A settlement was approved in January that reflected bondholders would recover about 2.7% to 3.4% of their funds, or about 3 cents on the dollar, according to court filings. The company’s former chairman was convicted of fraud by a Manhattan jury, and the Justice Department alleged he misappropriated more than $150 million from the company. He is scheduled to be sentenced in October.
GWG failed due to allegations of fraud and a weak structure. Long-life models have never failed, and the difference is important. The bonds were sold through an extensive network of brokerages, with commissions reported at 7% to 8%, an incentive that helped push illiquid speculative products to savers who didn’t know how much they owned. What this episode shows is narrower than fraud, but still relevant. Complex, illiquid mortality bets are sold to those looking for safety, wrapped in high yields, and when they unwind, ordinary investors may be left with little. Since then, the asset class has tightened its diligence standards, with independent life expectancy reports and stronger documentation now being taken as the baseline, but the lesson remains who ultimately bears the risk when these bets are unwound.
What people seeking income should ask
A high yield on death doesn’t necessarily mean it’s a bad investment, and for the right buyer, it truly is. Double-digit yields should raise questions, not solve them. Find out who behind the fund or policy creates the life expectancy estimates, whether the estimates are recent or independently created, and how returns change if the insured lives one to two years longer than predicted. Because that one sensitivity makes all the difference. Given that the industry is currently in the midst of a revision, ask how your position will be valued if the underwriting model is revised. And weigh the practical traps that the SEC warned about years ago and are still true today. Long wait times for payouts, steady premium outflows, thin liquidity, and loose regulation of the underwriters you’re guessing you’re buying from.
Whole life insurance payouts occur only if the insured dies, on a date that is not known to the investor in advance. The value of a bet depends on its date estimate, which is now increasingly done by machines.
