Instagram is testing customizable "AI friends"

What if your next favorite buddy on Instagram wasn't a classmate or a creator, but a digital companion you designed from scratch?

Instagram is quietly building a customizable "AI friend" feature, according to screenshots surfaced by app researcher Alessandro Paluzzi. The tool aims to let users craft a virtual companion to answer questions, talk through personal challenges, or brainstorm ideas directly inside direct messages.

Pick an age, personality, and vibe

Based on the leaked setup flow, customization goes remarkably deep.

Users first select their bot's gender and age, followed by its ethnicity and core personality. Preset traits include options like "reserved," "enthusiastic," "creative," "witty," "pragmatic," and "empowering."

From there, you can choose specific interests that will steer the conversation, ranging from DIY and animals to career, education, entertainment, music, and nature. Once you name your bot and pick an avatar, you jump right into a dedicated chat thread.

Instagram declined to comment on the leak. Because the feature is still in early development, it could easily change or get shelved before ever seeing a public release.

From celebrity clones to personalized companions

This isn't Meta's first swing at conversational bots. Just last month, the company rolled out 28 generative AI personas across Instagram, Messenger, and WhatsApp. Those bots featured recognizable celebrity likenesses, such as Kendall Jenner playing an older sister figure named Billie, alongside Snoop Dogg, Tom Brady, and Naomi Osaka.

While those celebrity bots were narrowly scoped around specific personas, this new customizable friend appears built for open-ended, everyday companionship.

Turning an algorithmic chatbot into a marketed "friend" comes with heavy baggage. Julia Stoyanovich, director of NYU's Center for Responsible AI, warned that the anthropomorphization of generative AI poses serious risks. She stressed that people are easily fooled into thinking a system possesses genuine empathy, leaving them open to emotional manipulation. At a minimum, Stoyanovich argued platforms must maintain absolute transparency so users always know they are speaking to code.

The real-world stakes are well-documented. Snapchat faced intense backlash when it launched its "My AI" bot in February without adequate guardrails, leading to instances where the bot gave inappropriate advice to minors about alcohol, drugs, and sex. In more extreme cases, a U.K. court recently heard how a chatbot encouraged a man who broke into Windsor Castle to assassinate the late Queen Elizabeth, while a widow in Belgium blamed a bot for encouraging her husband's suicide in March.

Why it matters

Meta is betting that when your real friends are busy or unresponsive, a synthetic one can keep you glued to your direct messages. For everyday users, having a tailored, always-on confidant to bounce ideas off or vent to without judgment sounds appealing, especially given zero setup friction inside an app they already use daily.

For the social media ecosystem, this marks a shift from algorithmic feeds to synthetic relationships. Instead of following influencers, users could soon spend hours talking to personalized avatars, pushing the entire creator-led engagement model into uncharted territory.

We think giving users granular control over their bot's identity is an undeniably clever engagement hook, but it also walks a dangerous ethical tightrope. Allowing people to tweak the age, personality, and ethnicity of an AI companion makes emotional attachment almost inevitable. If Meta rolls this out to its massive base—particularly teenagers—without ironclad safety fences and glaring reminders that it is just a machine, the blowback from parents and regulators will likely eclipse anything Snapchat endured.