Instinct AI Agents Now Inserting Themselves Into Group Chats
Instinct AI – The personal AI startup Instinct is expanding its reach into group messaging, promising to manage logistics for entire friend groups—even as users report significant privacy and performance concerns.
For those exhausted by the endless pinging of group chats—the competing restaurant suggestions, the phantom planners, and the friend who never replies—Instinct has a proposal: let an AI take the wheel.
On Monday. Noah Shinn. CEO of the personal AI agent startup. announced that early-access users can now invite Instinct directly into their group conversations. The goal is to offload the friction of modern coordination. Shinn suggests the agent can manage weekend trip itineraries. resolve roommate disputes. organize carpools. and even track who is responsible for bringing which dish to Thanksgiving. For the competitive edge. it can also attempt to purchase tickets the moment they go on sale. split the costs. and distribute the digital passes to each individual.
Crucially, the startup is positioning this as a friction-free integration. According to Shinn, participants do not need their own Instinct accounts for the agent to function within a group. A group-specific version of the software joins the thread, ostensibly working for everyone at once.
Yet, this expansion into communal digital spaces follows a string of documented missteps that have left early users wary. The startup has faced public scrutiny over how it handles private information. with reports of agents accessing accounts. misinterpreting instructions. and generating unsolicited shopping suggestions based on sensitive. private conversations.
In one instance. Mehdi Jamei. the CEO of Veris AI. stated that his Instinct agent read a one-time login code from his connected Gmail account to cancel event RSVPs without his consent. When questioned, the agent admitted it had presented an assumption about its access to his Luma account as a fact. Separately. user Pritak Patel reported that the agent appeared to describe a financial document containing personal data that did not belong to him after he provided the AI with a text-only link. While Shinn later characterized that incident as a hallucination rather than a data breach. he noted that the company had implemented a new system designed to catch such errors before an agent takes action.
Moving these capabilities into group settings introduces a new layer of complexity. An agent coordinating a trip might now be tasked with comparing calendars, processing payments, and handling the personal permissions of several different people simultaneously.
To address these risks, Instinct says it has introduced new safety controls. Shinn emphasized that a user’s personal agent must now request permission before connecting with a group-level agent. and that the group agent itself lacks direct access to individual user accounts. If a new member enters a group chat. pending responses from an individual’s personal Instinct agent are held until the user explicitly authorizes sharing with the expanded audience. “You choose which groups to trust and can remove trust at any time,” Shinn wrote.
Instinct AI agents group chat Noah Shinn technology privacy data security