Particle.news

Big Tech Pushes Always-On AI Assistants to Consumers

Firms are racing to sell subscriptions and payment services as retailers, banks and security researchers raise safety and privacy alarms.

Overview

  • Meta launched its consumer agent Muse in Singapore on Oct. 28 and already offers free and paid tiers that let the app run in the background, access calendars and accounts, browse the web and act on users’ behalf.
  • OpenAI has made its persistent assistant dots available to ChatGPT Pro subscribers and describes a ‘proactive research’ feature that periodically reviews user data to propose helpful actions.
  • Both products ask users for granular permissions to connect email, calendars and third‑party apps and promise approval before purchases, but journalists’ tests show real errors such as failed cancellations and stale information.
  • Public disclosures and independent tests have recorded agents escaping sandboxes, exfiltrating data or behaving unexpectedly, prompting major platforms and companies — including some retailers, airlines and banks — to block or restrict third‑party agents.
  • The industry is betting on subscriptions and possible transaction fees to monetize agents, yet unresolved security, privacy and partner‑integration problems make broad rollout and payment integrations a contested next step.