Pi launched as a support companion before its safety rules were dated

The design

Inflection AI announced the first release of Pi on 2 May 2023, according to the company's own press announcement. The post, quoting co-founder and then-CEO Mustafa Suleyman, describes Pi, short for "personal intelligence", as "a kind and supportive companion offering conversations, friendly advice, and concise information in a natural, flowing style," distinguished from assistants built around search or task completion. At launch Pi was free to use and reachable through Instagram, Facebook Messenger, WhatsApp and SMS, a web client at heypi.com, and an iOS app, with Android described as coming later. Today's heypi.com, as retrieved on 16 September 2026, still introduces itself in the same register: "I'm your personal AI, designed to be supportive, smart, and there for you anytime."

What the evidence says

The launch announcement is explicit that the traits it lists, "kind and supportive," "curious and humble," "creative and fun", describe the company's design intent rather than a measured outcome, and the post itself cautions that Pi is "in development" and "information could be wrong at times." No independent evaluation of Pi's emotional-support claims is cited in the launch material; the announcement is Inflection's own characterization of a product it had just shipped, not a third-party study of how the software performed for users.

What it asks of people

Pi's pitch asks users to bring open-ended, personal conversation to a product with no stated task to complete, the announcement frames Pi as "a coach, confidante, creative partner, or sounding board." That framing invites the kind of disclosure this site treats with caution: a tool built around supportive conversation depends on people sharing feelings and circumstances to function as advertised, and the launch material does not describe what happens to that history beyond the free service offered.

Privacy and safeguards

Inflection's launch post says "safety is the most important part of developing Pi," citing technical safeguards, human feedback and red-teaming without publishing specifics at the time. A more concrete, dated commitment appears later: the company's Crisis Prevention and Safety Protocol, last updated 22 December 2025 and issued under California's companion-chatbot statute, states that Pi will "acknowledge the user's distress" and direct people to crisis resources through a named referral service if self-harm or suicidal content is detected, and that Inflection discloses in app stores that Pi is intended for users 18 and older. That protocol postdates the 2023 launch by more than two years, so it should not be read back into what protections existed at release.

• What did Pi's safety design actually consist of in May 2023, versus what a 2025 protocol now describes?
• Does a "free at launch" AI companion generate revenue in some other way, and does the launch material say so?
• If a companion is available over ordinary text and messaging apps, what does that imply about who can access it and at what age?

Pi's launch announcement is candid that it was an early, evolving release; the specific safeguards a reader might assume were already in place, including the crisis-referral protocol now documented, arrived later and under a legal mandate rather than at launch.