Two major AI product launches this week are pushing the boundaries of what agents can do with your most sensitive personal data—but they take radically different approaches to privacy.
Kimi WebBridge, released by Beijing-based Moonshot AI, is a browser extension that lets AI agents interact with websites just like a human: clicking, typing, scrolling, and extracting information. Crucially, it runs entirely on the user's machine using the Chrome DevTools Protocol, so login sessions and page content never touch Moonshot's servers. The extension is agent-agnostic, officially supporting Kimi Code CLI, Claude Code, Cursor, Codex, and Hermes. It lets agents handle tasks like comparing prices across retailers, scanning job listings, or filling forms—all inside the user's existing Chrome or Edge window with cookies intact. The underlying model, Kimi K2.6, now tops the SWE-Bench Pro benchmark with a 58.6% score, ahead of GPT-5.4 and Claude Opus 4.6. The release follows a high-profile controversy in March 2026 when Cursor was found to have used Kimi's open-source model as the base for its Composer 2 without proper attribution.
In contrast, OpenAI just launched a personal finance feature for ChatGPT that connects to over 12,000 banks and credit cards via Plaid, giving the AI read-only access to balances, transactions, investments, and liabilities. Rolling out first to ChatGPT Pro subscribers ($200/month) in the U.S., the feature defaults to the GPT-5.5 Thinking model. Instead of generic savings tips, ChatGPT now analyzes actual spending patterns across categories and creates a personalized monthly budget with real dollar targets. OpenAI acquired two fintech startups—Roi and Hiro—to build this capability. The company stresses that Plaid handles authentication, data is synced according to the user's model training opt-in setting, and conversations are not used for training if the user has opted out. However, OpenAI explicitly states the service is not a financial advisor and carries no fiduciary duty.
The juxtaposition highlights a growing divergence in AI agent design: Kimi WebBridge keeps everything local, preserving privacy at the cost of cloud convenience, while OpenAI’s finance tool relies on centralized infrastructure to deliver a polished experience—raising questions about data security and trust. As both tools roll out to broader audiences, users will face a choice between local-first autonomy and the seamless, connected ecosystem that cloud-based AI can offer.
