Instinct AI assistant saves $550 but mishandles payment, raises security concerns
Instinct, an AI personal assistant, saved a Wired writer $550 and flagged a phishing email, but also mishandled a payment, highlighting security concerns. As this task-based AI model gains popularityโฆ
Instinct, an AIโpowered personal assistant that bills users per task, helped a Wired writer save $550, secure restaurant bookings and dodge a phishing email, but also cost $64 in a failed transaction and raised serious security concerns, the author reported on Tuesday.
The service launched early this year as part of a wave of โtaskโbasedโ AI agents that promise to handle everyday chores for a fee, positioning themselves as a cheaper alternative to subscriptionโbased chatbots. Consumers are drawn to the model because it charges only when the AI actually performs a job, a tempting proposition amid rising subscription fatigue. At the same time, regulators are scrutinising how these agents manage user data, after several highโprofile breaches involving AI tools that stored personal information in unsecured cloud buckets. Instinctโs mixed performance arrives as the market grapples with balancing convenience and privacy.
In the writerโs experience, Instinct automatically compared flight prices, booked a table at a downtown bistro and flagged a suspicious email that mimicked a bank notice. Those actions reportedly saved the author $550 in travel costs and prevented a potential loss of personal credentials. However, the AI also attempted to purchase a concert ticket on the writerโs behalf, misreading the payment link and charging $64 to a saved credit card. The transaction was later reversed, but the episode highlighted a lack of clear user controls and audit logs. Security researcher Dr. Maya Patel, who has examined several AI agents, warned that โany system that can act on your behalf must have robust authentication and transparent logging, or it becomes a vector for fraud.โ Instinctโs developers responded that they are rolling out twoโfactor verification for all financial actions and will release a dashboard showing every AIโinitiated request.
The episode underscores the tradeโoff between convenience and risk as AI agents become more embedded in daily life. Consumers may benefit from timeโsaving shortcuts, but they must weigh those gains against potential data exposure and erroneous actions. Instinct plans to publish a white paper on its security architecture next month and to invite thirdโparty audits. Regulators are expected to issue new guidance on AIโdriven financial transactions later this year, which could shape how services like Instinct operate and protect users.
Read Full Story at Wired โ


