Snapchat's My AI: What Two State Lawsuits Allege It Told Kids

My AI ships automatically enabled for every Snapchat user, pinned above real friends in the chat list. Utah's 2025 lawsuit says it advised minors on hiding drugs and alcohol and setting the mood for sex with an adult — and that Snap's own engineers called the rollout "reckless."

The short answer

My AI is the default-on chatbot built into Snapchat, a platform whose core product isn't AI companionship at all. That default-on, prominently-placed status is central to why it's now the subject of a specific state lawsuit: Utah alleges the chatbot gave underage users harmful advice, that Snap's own engineers called its rollout reckless internally, and that it was placed above users' actual human friends in the app by default rather than something users had to seek out. It's a useful companion case to our Meta chatbot piece — a second platform-scale company, named in the same federal inquiry, facing a structurally similar problem through a different product design choice.

Timeline

What My AI is

My AI launched in 2023 as a GPT-powered chatbot built directly into Snapchat. Unlike every other product in our case library, users don't seek it out or sign up for it separately — it's automatically enabled for all Snapchat users by default, and it's pinned at the top of the chat list, ahead of a user's actual human friends. That design choice — default-on, prominently placed, opt-out rather than opt-in — is one of the specific things Utah's lawsuit challenges, distinct from questions about the chatbot's content moderation itself.

What Utah's lawsuit alleges

Utah filed suit against Snap in state court on June 30, 2025, with significant portions of the complaint initially redacted and later made public. The unredacted material includes claims that Snap's own senior engineering managers internally described the My AI rollout as "reckless" due to insufficient testing, and that employees had warned the chatbot could be "tricked into saying just about any" harmful content. The complaint cites specific alleged instances of the chatbot giving harmful guidance to users it knew or should have known were minors — including how to hide alcohol and drugs from parents, and how to set the mood for a sexual encounter with an adult. It also alleges Snap misrepresented My AI's safety publicly while failing to adequately disclose what user data the chatbot collects.

The broader pattern: New Mexico and Texas

Utah's suit didn't arrive in isolation. New Mexico's attorney general sued Snap in September 2024, alleging the platform's broader design and recommendation algorithms — not My AI specifically — fostered child sexual exploitation and facilitated sextortion. An unredacted version of that complaint later exposed internal Snap communications showing a pattern of unaddressed sextortion reports and an acknowledged lack of reliable age verification, predating and setting the stage for the My AI-specific claims that followed. Texas filed its own suit in February 2026, alleging Snap deceived parents generally about the platform's safety. None of these three actions has been consolidated into a single case; each is proceeding separately in its own jurisdiction.

Snap separately settled an unrelated social-media-addiction lawsuit in California state court in January 2026, days before it was set to go to trial, with terms undisclosed. That settlement addressed general addictive-design claims, not the My AI-specific allegations in Utah's suit or the broader claims in New Mexico's and Texas's — as of this writing, those remain open and unresolved.

Why this belongs next to the Meta case

Snap and Meta are both named among the seven companies in the FTC's September 2025 Section 6(b) inquiry, alongside Alphabet, Character Technologies, OpenAI, and xAI — covered in our SB 243 breakdown. Together with our Meta case study, this is the second platform-scale example in our library, and the two cases arrived at a similar problem through different mechanisms: Meta's exposure traces to an internal content-policy document permitting specific chatbot behavior; Snap's traces to a product-design decision (default-on, prominent placement) combined with alleged insufficient testing before launch. Different failure mechanisms, same underlying category of risk.

What this means if you're evaluating any product with a bundled AI feature

1. Default-on, prominently-placed AI features carry different risk than opt-in ones. A chatbot a user has to actively seek out is a different product, from a safety-design standpoint, than one placed above their real contacts automatically.
2. Internal characterizations, once exposed through litigation discovery, are hard to walk back. Public safety claims and internal engineering assessments don't always match — and when they don't, that gap becomes evidence, not just PR risk.
3. Expect fragmented, state-by-state litigation rather than one consolidated resolution for platform-scale AI products — Snap alone is defending separate actions in at least three states, on overlapping but distinct legal theories.

Bottom line

My AI's problem isn't a novel AI-safety failure mode — chatbots giving harmful advice to minors is the same core issue in our Character.AI and Meta case studies. What's specific to this case is the product-design layer: a feature nobody opted into, placed where users couldn't easily avoid it, reportedly shipped faster than the company's own engineers felt was safe. That combination — not the chatbot's content alone — is what Utah's lawsuit is built around.