Retroactive Date

The earliest date on which a wrongful act can have occurred and still be covered under a claims-made policy, anchoring how far back prior acts are insured.

A retroactive date is the earliest date on which a wrongful act can have occurred and still be covered under a claims-made policy. Any act, error, or omission predating the retroactive date is outside coverage, regardless of when the resulting claim is made. The date is scheduled on the declarations page and is one of the most consequential negotiated terms on any claims-made form.

On a Generative AI Liability policy the retroactive date determines whether pre-policy AI deployments are insured. Standalone Gen AI forms typically scope the retroactive date to the first deployment date of the scheduled AI systems, which the insured discloses in the application. A retroactive date that reaches back to the actual launch of a production system is what makes the policy responsive to prior acts; a retroactive date set to policy inception leaves a gap for everything that came before.

The practical example: a deployer that integrated ChatGPT into a customer-facing product in 2024 and binds Gen AI Liability coverage in 2026 needs a retroactive date of 2024 (or earlier) for a 2026 claim arising out of a 2024 output to be covered. If the retroactive date is set to the 2026 inception, the claim falls outside the form even though the policy is in force when the claim is asserted.

Carriers vary in how they underwrite the retroactive date. Some grant full prior-acts coverage subject to a warranty that no claims or circumstances are known; others limit the retroactive date to a fixed window before policy inception; others tie it to the scheduled AI systems and refuse to cover systems not disclosed. Reading the schedule and matching it against actual deployment history is the underwriter's and broker's joint responsibility at binding.

Also known as

Prior Acts Date, Retro Date

Frequently asked

How is the retroactive date set on a Gen AI Liability policy?

Carriers typically scope the retroactive date to the first deployment date of the AI systems scheduled on the application. Disclosed systems with a longer production history get a longer prior-acts window; new systems may get a retroactive date matching policy inception. The applicant warrants that no known claims or circumstances exist as of the binding date. Systems not scheduled on the application generally fall outside the retroactive date entirely.

Does a retroactive date cover claims from before the policy was bound?

Yes, that is its function: a retroactive date of January 1, 2024 on a policy bound January 1, 2026 covers claims first made in 2026 arising out of wrongful acts occurring between January 1, 2024 and policy inception, subject to the policy's warranty that no known claims or circumstances existed at binding. Without a retroactive date reaching back, pre-policy acts are excluded even when the claim itself falls within the policy period.

What is the risk of a retroactive date set to policy inception?

Any wrongful act predating inception is uncovered, even if the resulting claim arrives during the policy period. For an AI deployer with a production system that has been live for two years before binding, a retroactive date set to inception strips coverage for the entire prior deployment history. A claim alleging that a 2024 output caused 2026 harm would fall outside the policy even though it is timely as a claims-made matter.

Related terms

generative AI liability insurance overview

General information, not legal or insurance advice.