On August 13, 2026, AIG disclosed $74 million in adverse prior-year development in its U.S. excess casualty book during second-quarter earnings, with accident years 2016 and 2023 cited as the primary contributors (earnings call transcript). AIG overall posted $145 million in net favorable prior-year development in the quarter, with the excess casualty charge offset by strong favorable emergence in workers’ compensation and property. The net number is good news for shareholders. The AY2016 item is a different kind of signal entirely.
A loss year from a decade ago is still generating reserve strengthening in Q2 2026. That is not an anomaly. It is a data point about how long the effective tail on commercial casualty actually runs, and it has direct implications for any self-insured whose actuarial opinion uses a development triangle shorter than ten years.
What AY2016 developing in 2026 means
Management framed the AY2016 adjustment as “prudent alignment with the level of conservatism in subsequent years.” That language is specific. AIG is not describing unexpected frequency or severity in ten-year-old claims. It is describing a decision to revise upward where AY2016 ultimately lands, based on a comparison with the conservatism baked into newer accident years. The practical meaning: the company’s view of ultimate loss on 2016 business moved in Q2 2026, after ten full development periods.
Long-tail casualty takes that long because cases work through mediation, trial, and appeal over years. Litigation funding, which became a significant market force after 2018, has extended resolution timelines further. A claim filed in 2016 under a general liability or commercial umbrella program may still be in active litigation in 2026. If the carrier above the self-insured attachment is still revising AY2016 upward, the primary retained layer for those same accident years may be carrying the same trajectory.
The AY2023 finding is a different problem
The concurrent AY2023 adverse result deserves separate attention. Management confirmed “there’s no change in frequency or severity” on the 2023 book. The strengthening was a deliberate decision to bring AY2023 “in line with the level of prudence reflected in 2024 and 2025.” That framing means initial case reserves on 2023 claims were set below the level management now believes is appropriate for a book of that maturity. AY2023 is three years old. Strengthening it now, before it is remotely mature, signals that the early case reserve adequacy on recent claims is not where it needs to be.
For self-insureds, the AY2023 signal is diagnostic: if the excess carrier is revising upward at three years of development, the incurred development on the primary retained layer for that same accident year is likely tracking in the same direction, and the pick your actuary used when the year was fresh may be stale.
Who it affects
Self-insured employers with general liability SIRs above $250,000, single-parent captives writing umbrella or excess casualty limits, and group captives or risk retention groups with significant GL or commercial umbrella exposure. The AY2016 tail length finding is not specific to AIG’s book. The pattern is confirmed by The Hartford, which posted $116 million in adverse general liability development and $26 million in commercial auto in Q2 2026, and by Everest Group, which strengthened North America treaty casualty reserves by roughly $200 million across most accident years in the same quarter. Three major carriers with deep casualty triangles are simultaneously developing adversely on recent accident years and still adjusting accident years from a decade ago. That convergence is not a coincidence.
Reserve mechanism: tail factor truncation
The chain-ladder method translates observed loss emergence into ultimate loss estimates by applying development factors from a triangle of historical data. When that triangle ends at seven years, the tail factor compresses all remaining liability beyond year seven into a single multiplier. If the actual tail runs to year ten or beyond, as the AY2016 data suggests, that multiplier is understated and IBNR is systematically low on every open accident year in the program.
Tail factors for smaller self-insured programs are typically derived from ISO development patterns, NCCI benchmarks, or the reinsurer’s aggregate book. All of those sources are built from the same historical casualty emergence that AIG’s AY2016 finding is now revising upward. Programs with fewer than ten years of credible own-book claims data are the most exposed: there is no internal history to flag the truncation, and the external benchmark carries the same understatement.
RPS’s Q2 2026 umbrella and excess market update documents loss trends in the book at 12-15%, with rising indemnity, defense costs, and settlements all contributing. At that trend rate, a tail factor shortfall of even a few points compounds across four to six open accident years in a typical casualty self-insured’s portfolio.
What this means for your next review
Ask your actuarial consultant how many accident years are included in the development triangle for general liability and umbrella lines. If the answer is fewer than ten, ask specifically how the tail factor for those lines is derived and whether it was benchmarked against post-2018 industry emergence patterns or against the pre-social-inflation baseline. The difference matters: industry triangles compiled before 2019 do not reflect the longer resolution timelines that litigation funding has introduced.
On the AY2023 question: ask whether the current case reserve adequacy assumption for your 2023 claims is consistent with what carriers at the excess layer are now booking. If your program’s AY2023 is still anchored to the initial loss picks set at the start of that year, the incurred triangle for that cohort may already be showing the same upward drift that AIG recognized in Q2.