Iris

Support at three in the morning, in the thread people already use at three in the morning. Iris describes itself as a non-clinical companion rather than as therapy, which is the distinction that matters and the one this listing records rather than assesses. The number is published and answers without an account.

Two things in the policy are worth having in front of you before the first message. It states that Iris includes automated features meant to detect crisis-related content and show mental health resources, and then says those systems are not guaranteed to catch everything and that you should not rely on Iris to recognise or respond to an emergency. It also lists session monitoring tools among the parties that receive data, to support the quality and safety of AI interactions, which means somebody or something other than the model may read the thread.

Its AI providers are barred from retaining the data or training on it, and none of them is named. The policy carries an unfilled placeholder where the effective date should be, so there is no date to record and no way to tell when any of this last changed.

Who is behind it is stated in one line on a page most readers will not reach: Iris is created by a Flagship Pioneering company, the venture firm best known for founding Moderna. Nothing else about the company, the founders or the money is published, and neither is a price.

Agentcard, a wallet built for agents, names Iris among the products it is trusted by on its own front page, which is the only public statement about how Iris would pay for anything; nothing on Iris's pages says so, and the capability row stays as Iris states it.

One word in its privacy policy changed between 10 and 17 September 2026. The sentence that said "AI companion services, including IrisChat, may not be suitable for minors" now says "AI thought partner services". Companion is the word California's SB 243 is written around, and the listing records the change without saying why it was made.